Supplementary Appendix

Transcription

1 Supplementary Appendix This appendix has been provided by the authors to give readers additional information about their work. Supplement to: Goldstone AB, Chiu P, Baiocchi M, et al. Mechanical or biologic prostheses for aortic-valve and mitral-valve replacement. N Engl J Med 2017;377: DOI: /NEJMoa

2 Mechanical or Biological Prostheses for Aortic- and Mitral-Valve Replacement Supplementary Appendix Andrew B. Goldstone, MD, PhD,* Peter Chiu, MD, MS,* Michael Baiocchi, PhD, Bharathi Lingala, PhD, William L. Patrick, BS, Michael P. Fischbein, MD, PhD, Y. Joseph Woo, MD Table of Contents Page Number Methods Author Contributions 5 Methods Funding Sources 5 Methods Use of Inverse Probability Weighting 5 Methods Lack of Access to Strong Instrumental Variables 6 Methods Restricted Mean Survival Time 7 Methods Power Calculations 8 Methods Internal Validation Study 8 Methods Sensitivity to Misclassification of Valve Type 9 Methods Sensitivity to Unmeasured Confounding 10 Methods Secular Trends 10 Methods Statistical Packages 11 Figure S1 Distribution of Propensity Scores, Stratified by Age Group and 13 Type for Aortic-Valve Replacement Figure S2 Distribution of Propensity Scores, Stratified by Age Group and 14 Type for Mitral-Valve Replacement Figure S3 Distribution of Stabilized Weights, Stratified by Age Group and 15 Type for Aortic-Valve Replacement Figure S4 Distribution of Stabilized Weights, Stratified by Age Group and 16 Type for Mitral-Valve Replacement Figure S5 Patient Selection Flow Diagram for Aortic-Valve Replacement 17 Figure S6 Patient Selection Flow Diagram for Mitral-Valve Replacement 18 Figure S7 Mitral-Valve Replacement vs. Repair Cases Per Year, Stratified by 19 Age Group Figure S8 Aortic-Valve Utilization Per Year, Stratified by Age Group 20 Figure S9 Mitral-Valve Utilization Per Year, Stratified by Age Group 21 Figure S10 Net Survival Penalty or Benefit of Bioprosthetic Aortic-Valves Over 22 Time (Additional Area Under the Survival Curve) Figure S11 Mortality after Aortic-Valve Replacement with Mechanical or 23 Biological Prostheses Stratified by Time Period Ages Years Figure S12 Mortality after Aortic-Valve Replacement with Mechanical or 24 Biological Prostheses Stratified by Time Period Ages Years Figure S13 Cumulative Incidence of Stroke after Aortic-Valve Replacement, 25 Stratified by Age and Type Figure S14 Cumulative Incidence of Stroke after Aortic-Valve Replacement, 26 Stratified by Time Period and Type Ages Years Figure S15 Cumulative Incidence of Stroke after Aortic-Valve Replacement, 27 Stratified by Time Period and Type Ages Years Figure S16 Cumulative Incidence of Bleeding after Aortic-Valve Replacement, 28 1

3 Stratified by Age and Type Figure S17 Cumulative Incidence of Bleeding after Aortic-Valve Replacement, 29 Stratified by Time Period and Type Ages Years Figure S18 Cumulative Incidence of Bleeding after Aortic-Valve Replacement, 30 Stratified by Time Period and Type Ages Years Figure S19 Cumulative Incidence of Reoperation after Aortic-Valve 31 Replacement, Stratified by Age and Type Figure S20 Cumulative Incidence of Reoperation after Aortic-Valve 32 Replacement, Stratified by Time Period and Type Ages Years Figure S21 Cumulative Incidence of Reoperation after Aortic-Valve 33 Replacement, Stratified by Time Period and Type Ages Years Figure S22 Net Survival Penalty or Benefit of Bioprosthetic Mitral-Valves Over 34 Time (Additional Area Under the Survival Curve) Figure S23 Mortality After Mitral-Valve Replacement with Mechanical or 35 Biologic Prostheses Stratified by Decade of Age Figure S24 Net Survival Penalty or Benefit of Bioprosthetic Mitral-Valves Over 36 Time (Additional Area Under the Survival Curve) in Patients and Years Old Figure S25 Mortality After Mitral-Valve Replacement with Mechanical or 37 Biological Prostheses Stratified by Time Period Ages Years Figure S26 Cumulative Incidence of Stroke after Mitral-Valve Replacement, 38 Stratified by Age and Type Figure S27 Cumulative Incidence of Stroke after Mitral-Valve Replacement, 39 Stratified by Time Period and Type Ages Years Figure S28 Cumulative Incidence of Bleeding after Mitral-Valve Replacement, 40 Stratified by Age and Type Figure S29 Cumulative Incidence of Bleeding after Mitral-Valve Replacement, 41 Stratified by Time Period and Type Ages Years Figure S30 Cumulative Incidence of Reoperation after Mitral-Valve 42 Replacement, Stratified by Age and Type Figure S31 Cumulative Incidence of Reoperation after Mitral-Valve 43 Replacement, Stratified by Time Period and Type Ages Years Table S1 Exclusion Criteria for Aortic-Valve Replacement Cohort 44 Table S2 Exclusion Criteria for Mitral-Valve Replacement Cohort 45 Table S3 Definitions of Baseline Characteristics 46 Table S4 Definitions of Study Endpoints 52 Table S5 Models for Estimating Propensity Scores for Aortic-Valve 54 Replacement Table S6 Models for Estimating Propensity Scores for Mitral-Valve 55 Replacement Table S7 Baseline and Operative Characteristics of Patients Years Old 56 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S8 Baseline and Operative Characteristics of Patients Years Old 57 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S9 Baseline and Operative Characteristics of Patients Years Old 58 2

4 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S10 Baseline and Operative Characteristics of Patients Years Old 59 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S11 Baseline and Operative Characteristics of Patients Years Old 60 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S12 Baseline and Operative Characteristics of Patients Years Old 61 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S13 Baseline and Operative Characteristics of Patients Years Old 62 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S14 Baseline and Operative Characteristics of Patients Years Old 63 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S15 Baseline and Operative Characteristics of Patients Years Old 64 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S16 Baseline and Operative Characteristics of Patients Years Old 65 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S17 Baseline and Operative Characteristics of Patients Years Old 66 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S18 Baseline and Operative Characteristics of Patients Years Old 67 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S19 Baseline and Operative Characteristics of Patients Years Old 68 Before and After Inverse Probability Weighting for Aortic-Valve Replacement Cohort Table S20 Baseline and Operative Characteristics of Patients Years Old 69 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S21 Baseline and Operative Characteristics of Patients Years Old 70 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S22 Baseline and Operative Characteristics of Patients Years Old 71 Before and After Inverse Probability Weighting for Mitral-Valve Replacement Cohort Table S23 Between-Group 30-Day Mortality Differences for Recipients of 72 Biological and Mechanical Prostheses Table S24 Sensitivity Analysis Against Misclassification of Type 79 Table S25 Sensitivity Analysis Against Unmeasured Confounding for Aortic- 80 Valve Replacement Cohort Table S26 Subdistribution Hazard of Stroke, Bleeding, or Reoperation among 73 Biological versus Mechanical (Reference) Aortic-Valve Prostheses with Death as a Competing Risk Table S27 Hazard of Death for the Time-Varying Complications of Stroke and 74 3

5 Bleeding after Aortic-Valve Replacement Table S28 Between-Group Mortality Differences for Recipients of Mechanical 75 (Reference) and Bioprosthetic Mitral Valves in Patients and Years Old Table S29 Sensitivity Analysis Against Unmeasured Confounding for Mitral- 81 Valve Replacement Cohort Table S30 Subdistribution Hazard of Stroke, Bleeding, or Reoperation among 76 Biological versus Mechanical (Reference) Mitral-Valve Prostheses with Death as a Competing Risk Table S31 Hazard of Death for the Time-Varying Complications of Stroke and 77 Bleeding after Mitral-Valve Replacement Table S32 Positive Predictive Value of Valve Replacement ICD9-CM Codes Stratified by Age (Internal Validation) 78 References Supplementary Appendix References 82 4

6 METHODS Author Contributions Andrew B. Goldstone, Peter Chiu, Michael Baiocchi, and Y. Joseph Woo designed the study. Andrew B. Goldstone and William L. Patrick reviewed and compiled ICD9-CM codes for inclusion criteria, exclusion criteria, and study endpoints. The data were provided by the California Office of Statewide Health Planning and Development (OSHPD). Bharathi Lingala output the data into an analyzable format. Andrew B. Goldstone and Peter Chiu analyzed the data. All authors vouch for the validity of the data and analysis. Andrew B. Goldstone and Peter Chiu wrote the first draft of the manuscript. All authors substantially contributed to discussion of content and revision of the manuscript prior to submission. All authors were involved in the decision to and agreed to publish the manuscript. Funding Sources This work was conducted with support from a TL1 Clinical Research Training Program and KL2 Mentored Career Development Award of the Stanford Clinical and Translational Science Award to Spectrum (ABG - NIH TL1 TR ; PC - NIH KL2 TR ), as well as a grant from the Agency for Healthcare Research and Quality (MB - KHS022192A). Statistical Analysis Use of Inverse Probability Weighting A frequently employed technique in clinical research is propensity score pair matching, which creates matches between two patients who received different treatments but who had similar a probability of receiving the same treatment. In this design, many patients are excluded 5

7 due to lack of an appropriate match. The generalizability and interpretability of results can be limited when patients are excluded in this manner. Our goal was to investigate the effect of treating all patients with biological prostheses (average treatment effect); this is the effect estimate of a change in guidelines. Two techniques for the estimation of the average treatment effect were considered: inverse probability weighting and full matching. Inverse probability weighting incorporates the entire study population and weights patients by the inverse of the probability of receiving the observed treatment to create balanced pseudopopulations. Like inverse probability weighting, full matching incorporates the entire study population but does so by creating matched sets of variable size. This is in contrast to the more typical pair matching technique, which limits matched sets to 2 observations. Full matching is rarely used, and its use has primarily been restricted to binary or continuous outcomes. 1 Therefore, we chose to use inverse probability weighting to address measured confounding. Lack of Access to Strong Instrumental Variables The instrumental variable method was developed to control for unmeasured confounders. An instrumental variable method exploits a variable that influences treatment assignment with the following two conditions: it is independent of unmeasured confounders, and it does not directly affect outcome except through its influence on treatment assignment. 2 For this analysis, we explored the possibility of using institution as a preference-based instrumental variable (the specific surgeon is not available through OSHPD). Preference-based instrumental variables are particularly well-suited for observational studies of surgical treatments because: 1) the specific treatment a patient receives is often determined by the preferences of the surgeon performing the operation, and 2) for routine operations, patients are typically referred to surgeons in a manner that is not informed by severity or complexity of illness. In the case of valve replacement, a patient who received a biological prosthesis might have received a mechanical prosthesis had he/she been referred to a different surgeon. Institutions, however, 6

8 often have several operating surgeons, each with their own preferences, training, and referral pattern. For this reason, institution was a weak instrument in our study population, particularly in the primary age groups of interest (F-statistic approximately 1 in first stage least squares model; calculated in R with ivmodel package 3 ). We do believe, however, that individual surgeon is theoretically a strong preference-based instrument and propose future analyses be performed if such surgeon-level data is available. Restricted Mean Survival Time Time-to-event analyses typically employ Cox proportional hazards models. When comparing two groups in this fashion, an assumption is made that the ratio of the two hazard functions for the groups remains constant over time (e.g. proportional hazards). The Cox model is then used to estimate the unknown constant hazard ratio parameter. When there is concern, or evidence, of non-proportional hazards (e.g. the hazard ratio changes over time), the estimated hazard ratio may not be a meaningful measure of the between-group survival difference. 4 Despite this issue, the Cochrane handbook instructs investigators to report hazard ratios to quantify between-group differences in survival, a recommendation supported by the CONSORT guidelines. 5,6 For this reason, we report hazard ratios in our manuscript, but we also chose to report the restricted mean survival time (RMST). 7,8 The RMST is the population s average duration of event-free survival experienced over a pre-specified follow-up period and is calculated as the area under the survival curve. In our study, we chose 15 years as our followup duration because a modest number of patients remained present in the risk set, and it preceded the last observed event. A benefit of the RMST is that it does not rely on the proportional hazards assumption. Once the area under the survival curve is calculated, the difference in RMST between groups and the ratio of the two numbers provide useful estimates of the treatment effect on overall survival. The 15-year RMST difference is interpreted as the average number of additional days gained by the treatment arm. The RMST ratio quantifies the 7

9 percent of life gained by the treatment arm compared to the control arm. Similarly, the area above the survival curve corresponds to the number of days of life lost until the designated time point (e.g. 15 years), and is known as the restricted mean time lost. These model-free estimates of survival benefit also may have better power to detect between-group survival differences compared with the log-rank test in the setting of non-proportional hazards. 9 Power Calculations The study was designed to achieve a power of at least 85% with an alpha-level of 0.05 to detect a between-group hazard ratio (HR) of 1.15 for 15-year mortality in patients 55-to-64 years old who underwent aortic-valve replacement, and in patients 50-to-69 years old who underwent mitral-valve replacement. 10 We assumed a ratio of 2:1 mechanical to biological prostheses, a median survival of 15 years, and an 18-year accrual period with no additional follow-up. 11,12 Among patients 45-to-54 years old undergoing aortic-valve replacement, we had 86% power to detect a hazard ratio of 1.25 (assuming a ratio of 2:1 mechanical to biological prostheses and a median survival of 20 years). Among patients 40-to-49 years old undergoing mitral-valve replacement, we had 86% power to detect a hazard ratio of 1.6 (assuming a ratio of 5:1 mechanical to biological prostheses and a median survival of 25 years). Among patients 70- to-79 years old, we had 96% power to detect a hazard ratio of 1.15 (assuming a ratio of 1:1 mechanical to biological prostheses and a median survival of 8 years). Internal Validation Study Although the description of each ICD9-CM procedure code distinguishes between implantation of a mechanical or a biological prosthesis, the accuracy of these codes in practice has never been described. We conducted an internal validation study to determine the accuracy of each procedure code. All patients who underwent isolated surgical aortic-valve replacement or mitral-valve replacement (the latter with or without coronary bypass, tricuspid-valve repair, or 8

10 ablation of atrial fibrillation) between 1993 and 2014 at Stanford University Medical Center were queried (N=1,067 for aortic-valve replacement, N=503 for mitral-valve replacement). The valve type received by the patient was then compared with the ICD9-CM procedure code billed for that visit. The sensitivity of and (aortic-valve replacement or mitral-valve replacement with a mechanical valve, respectively) was 80% and 88%. The sensitivity of and (aortic-valve replacement or mitral-valve replacement with a biological valve, respectively) was 75% and 76%. Theoretically, the specificity of each procedure code (e.g. the proportion of patients seen at Stanford who did not undergo valve replacement and were not billed for a valve replacement) is effectively 100%. As it applies to validation of administrative data, the positive predictive value (PPV) of each code may be more important than the sensitivity or specificity (Table S32). We found that misclassification of valve type varied with age and valve position. The PPV for was >90% below age 65 years, but <50% above age 65 years (embedded table). The PPV for was >95% below age 70 years, 70% between 70 and 79 years old, and 0% above 80 years old. Therefore, we excluded patients in our study who were 65 years old at the time of index aortic-valve replacement because the misclassification of valve type was substantial. Similarly, we excluded patients 80 years old at the time of index mitral-valve replacement. Sensitivity to Misclassification of Valve Type We performed a sensitivity analysis to estimate whether the overall results of our study were robust to misclassification of valve type. Using the Monte Carlo method, we randomly reassigned X% of mechanical prosthesis recipients to a biological prosthesis and Y% of biological prosthesis recipients to a mechanical prosthesis, where X% and Y% were 1 minus the PPV of each ICD9-CM procedure code. The distribution of the hazard ratio for a biological prosthesis was then determined with bootstrap resampling and re-estimation of stabilized 9

11 inverse probability weights. A one-sided 95% sensitivity interval was obtained over all 40,000 Monte Carlo and bootstrap iterations (Table S24). Sensitivity to Unmeasured Confounding We estimated outcomes under different levels of residual confounding using the sensitivity framework proposed by Klungsøyr and colleagues. 13 This analysis is designed to estimate the amount of residual confounding necessary to modify a given result from a Cox proportional hazards model which includes inverse probability weights. To understand the results of this sensitivity analysis, one must first consider the amount of baseline variation that is already accounted for with inverse probability weighting. We balanced our treatment groups on >40 baseline covariates for our primary analyses. Therefore, when conducting the sensitivity analysis, we ask: what would be the impact on our estimate of the treatment effect if we could sort two nominally identical patients (in >40 covariates) that was alpha times more likely to experience the outcome, i.e. death during follow-up, to one treatment rather than the other? Phrased in more usual (though admittedly less precise) language: if two patients with the same 42 baseline characteristics were standing in front of a surgeon, is it believable that the surgeon could still detect an alpha level difference in probability of death and consequently assign that patient to a particular treatment because of it? If it is plausible that surgeons are always able to route the sicker patients to receive bioprosthetic valves (despite having identical risk profiles over the 42 baseline characteristics), then there is concern for residual confounding. We used a constant function for the theoretical unmeasured confounder, and results are presented in Tables S25 and S29. Secular Trends Given the duration of time under study, a major concern of ours was the possibility of a secular trend confounding the association of valve type and mortality. To test for an era effect, 10

12 we performed a sensitivity analysis by separating the study period into 6-year time intervals. We then used inverse probability weighting with stabilized weights as described previously to limit the effect of confounding by indication. Weighted Cox proportional hazards regression with a robust variance estimator was used to assess whether prosthesis type influenced mortality. A weighted form of the sub-distribution hazard described by Fine and Gray was then used to evaluate incidence of stroke, bleeding, and reoperation with death as a competing risk. Our general inference did not appear to be markedly influenced by a secular trend given that many point estimates were in line with our main analysis (Figures S11, S12, S14, S15, S17, S18, S20, S21, S25, S27, S29, S31). However, two significant issues arose with this sensitivity analysis. Firstly, separating the data into 6-year time intervals drastically reduced the number of patients within each comparison and in turn drastically reduced the power. Secondly, our ability to detect differences in the more recent eras was severely reduced by lack of follow-up. Thirdly, risks of complications change over time. For example, patients with 6 years of follow-up are at risk of the complications of anticoagulation associated with mechanical valves, but not at significant risk of valve deterioration associated with bioprosthetic valves. Given these considerations, when divided into six-year time intervals, the benefit afforded by a mechanical mitral valve in patients years old was evident in only those with at least 12 years of followup; but this reduction in mortality was not statistically significant after correcting for multiple comparisons (Figure S25). It is possible that later generation prostheses are superior to earlier generation prostheses, but further studies are needed to determine whether this is the case once adequate follow-up has accumulated. Statistical Packages Data was prepared for analysis in SAS (SAS Institute Inc., Cary, NC) and statistical analyses were performed in R version (R Foundation, Vienna, Austria). The survey package 14 was used to compare baseline characteristics, 30-day mortality, and other 11

13 perioperative outcomes. All-cause mortality and competing risks analyses were performed with the survival package. 15 Weighted Cox proportional hazards regression, with or without baseline characteristics forced into the model, was used with an independence covariance matrix and robust variance estimator to estimate the marginal and conditional effects of valve type on allcause mortality. An independence working correlation matrix is recommended in the setting of inverse probability weighting 16 and is employed in the survival package. Because operations are nested within hospitals, we also used the coxme package 17 to perform mixed-effects Cox proportional hazards models of the weighted pseudopopulation with hospital as a random baseline hazard. The age-dependent effect of prosthesis type on the hazard of death was also created with the survival package. Patients within OSHPD who were 40-to-45 years old at the time of isolated aortic-valve replacement were included for modeling the spline in order to stabilize the estimate at the extremes of age. RMST differences, RMST ratios, and RMTL ratios were calculated and compared between groups with the survrm2 package. 18 In an effort to improve transparency in clinical research, the authors are willing to provide the code used to perform all calculations within the manuscript on request. 12

14 Figure S1. Distribution of Propensity Scores, Stratified by Age Group and Type for Aortic-Valve Replacement A Years of Age B Years of Age Mechanical: median 0.28 (IQR 0.19 to 0.40), range 0.00 to 0.73 Biological: median 0.39 (IQR 0.29 to 0.49), range 0.08 to 0.86 Mechanical: median 0.36 (IQR 0.23 to 0.51), range 0.05 to 0.84 Biological: median 0.50 (IQR 0.38 to 0.59), range 0.11 to

15 Figure S2. Distribution of Propensity Scores, Stratified by Age Group and Type for Mitral-Valve Replacement A Years of Age B Years of Age Mechanical: median 0.12 (IQR 0.07 to 0.22), range 0.00 to 0.77 Biological: median 0.26 (IQR 0.16 to 0.36), range 0.02 to 0.91 C Years of Age Mechanical: median 0.21 (IQR 0.12 to 0.34), range 0.02 to 0.91 Biological: median 0.39 (IQR 0.26 to 0.50), range 0.03 to 0.90 D Years of Age Mechanical: median 0.16 (IQR 0.08 to 0.28), range 0.00 to 0.91 Biological: median 0.31 (IQR 0.20 to 0.42), range 0.03 to 0.93 Mechanical: median 0.26 (IQR 0.15 to 0.40), range 0.00 to 0.87 Biological: median 0.43 (IQR 0.31 to 0.54), range 0.06 to 0.99 E Years of Age Mechanical: median 0.42 (IQR 0.33 to 0.58), range 0.11 to 0.90 Biological: median 0.58 (IQR 0.43 to 0.66), range 0.13 to

16 Figure S3. Distribution of Stabilized Weights, Stratified by Age Group and Type for Aortic-Valve Replacement A Years of Age B Years of Age Mechanical: median 0.93 (IQR: 0.83 to 1.12), range: 0.67 to 2.49 Biological: median 0.84 (IQR: 0.68 to 1.16), range: 0.38 to 3.94 Mechanical: median 0.91 (IQR: 0.76 to 1.18), range: 0.61 to 3.63 Biological: median 0.83 (IQR: 0.71 to 1.09), range: 0.42 to

17 Figure S4. Distribution of Stabilized Weights, Stratified by Age Group and Type for Mitral-Valve Replacement A Years of Age B Years of Age Mechanical: median 0.94 (IQR 0.88 to 1.05), range 0.82 to 3.58 Biological: median 0.67 (IQR 0.50 to 1.12), range 0.20 to 8.35 C Years of Age Mechanical: median 0.90 (IQR 0.81 to 1.09), range 0.73 to 8.15 Biological: median 0.74 (IQR 0.57 to 1.08), range 0.32 to 9.77 D Years of Age Mechanical: median 0.92 (IQR 0.85 to 1.07), range 0.78 to 8.61 Biological: median 0.72 (IQR 0.53 to 1.08), range 0.24 to 7.88 Mechanical: median 0.90 (IQR 0.79 to 1.12), range 0.67 to 5.04 Biological: median 0.77 (IQR 0.62 to 1.07), range 0.33 to 5.45 E Years of Age Mechanical: median 0.86 (IQR 0.74 to 1.19), range 0.56 to 5.11 Biological: median 0.86 (IQR 0.75 to 1.16), range 0.55 to

18 Figure S5. Patient Selection Flow Diagram for Aortic-Valve Replacement OSHPD Claims for Isolated Aortic-Valve Replacement N = 45,639 Exclusions (N=35,697) Prior AVR: 811 Prior cardiovascular operation: 6,353 More than 1 operation at the index hospitalization: 114 Age <45: 1,092 Age >65: 27,884 Surgery before 1/96 or after 12/13: 700 Residency outside of California: 166 Duplicate RLN: 4 Same day reoperation: 5 Overall Study Population N = 9,942 Mechanical N=6,097 Biological N= 3,845 17

19 Figure S6. Patient Selection Flow Diagram for Mitral-Valve Replacement OSHPD Claims for Mitral-Valve Replacement Exclusions (N=22,928) Concomitant procedures: 13,109 Prior mitral-valve surgery: 814 Prior other valve or heart surgery: 3,969 Age <40 years: 1,309 Age 80 years: 2,338 Residency outside of California: 367 Surgery before 1/96 or after 12/13: 940 Coded as receiving both bioprosthetic and mechanical mitral valve: 34 Coded as receiving two valves in same hospitalization: 47 Missing demographic data: 1 N=38,431 Overall Study Population N=15,503 Mechanical Biological N=9,982 N=5,521 18

20 Figure S7. Mitral-Valve Replacement vs. Repair Cases Per Year, Stratified By Age Group A Years of Age B Years of Age C Years of Age D Years of Age E Years of Age F Entire Mitral-Valve Population Maroon color represents overlap of mitral valve replacement and repair. 19

21 Figure S8. Aortic-Valve Utilization Per Year, Stratified By Age Group A Years of Age B Years of Age C All Aortic-Valve Replacements Maroon color represents overlap of mitral valve replacement and repair. 20

22 Figure S9. Mitral-Valve Utilization Per Year, Stratified By Age Group A Years of Age B Years of Age C Years of Age D Years of Age E Years of Age F All Mitral-Valve Replacements Maroon color represents overlap of mechanical and biological prostheses. 21

23 Figure S10. Net Survival Penalty or Benefit of Bioprosthetic Aortic Valves Over Time (Additional Area Under the Survival Curve) A Years of Age B Years of Age Red circles represent point estimates of the difference in restricted mean survival time between bioprosthetic and mechanical valve recipients each year after surgery. Dashed lines are 95% confidence intervals. RMST, restricted mean survival time 22

24 Figure S11. Mortality after Aortic-Valve Replacement with Mechanical or Biological Prostheses Stratified by Time Period Ages Years A 1996 to 2001 Hazard Ratio 1.13 (95% CI 0.87 to 1.48) B 2002 to 2007 Hazard Ratio 1.13 (95% CI 0.85 to 1.51) C 2008 to 2013 Hazard Ratio 1.41 (95% CI 0.89 to 2.23) 23

25 Figure S12. Mortality after Aortic-Valve Replacement with Mechanical or Biological Prostheses Stratified by Time Period Ages Years A 1996 to 2001 Hazard Ratio 1.12 (95% CI 0.93 to 1.34) B 2002 to 2007 Hazard Ratio 1.04 (95% CI 0.86 to 1.27) C 2008 to 2013 Hazard Ratio 0.99 (95% CI 0.70 to 1.41) 24

26 Figure S13. Cumulative Incidence of Stroke after Aortic-Valve Replacement, Stratified by Age and Type A Stroke in Patients Years Old Hazard Ratio 0.64 (95% CI 0.46 to 0.86) B Stroke in Patients Years Old Hazard Ratio 0.92 (95% CI 0.73 to 1.13) 25

27 Figure S14. Cumulative Incidence of Stroke after Aortic-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 0.64 (95% CI 0.37 to 1.10) B 2002 to 2007 Hazard Ratio 0.63 (95% CI 0.41 to 0.97) C 2008 to 2013 Hazard Ratio 0.70 (95% CI 0.37 to 1.34) 26

28 Figure S15. Cumulative Incidence of Stroke after Aortic-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 0.83 (95% CI 0.57 to 1.22) B 2002 to 2007 Hazard Ratio 0.97 (95% CI 0.71 to 1.33) C 2008 to 2013 Hazard Ratio 1.27 (95% CI 0.78 to 2.06) 27

29 Figure S16. Cumulative Incidence of Bleeding after Aortic-Valve Replacement, Stratified by Age and Type A Bleeding in Patients Years Old Hazard Ratio 0.63 (95% CI 0.51 to 0.75) B Bleeding in Patients Years Old Hazard Ratio 0.66 (95% CI 0.58 to 0.75) 28

30 Figure S17. Cumulative Incidence of Bleeding after Aortic-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 0.62 (95% CI 0.44 to 0.87) B 2002 to 2007 Hazard Ratio 0.60 (95% CI 0.45 to 0.79) C 2008 to 2013 Hazard Ratio 0.52 (95% CI 0.36 to 0.76) 29

31 Figure S18. Cumulative Incidence of Bleeding after Aortic-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 0.76 (95% CI 0.60 to 0.97) B 2002 to 2007 Hazard Ratio 0.62 (95% CI 0.51 to 0.75) C 2008 to 2013 Hazard Ratio 0.62 (95% CI 0.48 to 0.81) 30

32 Figure S19. Cumulative Incidence of Reoperation after Aortic-Valve Replacement, Stratified by Age and Type A Reoperation in Patients Years Old Hazard Ratio 2.60 (95% CI 1.91 to 3.40) B Reoperation in Patients Years Old Hazard Ratio 2.46 (95% CI 1.93 to 3.20) 31

33 Figure S20. Cumulative Incidence of Reoperation after Aortic-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 2.97 (95% CI 2.04 to 4.33) B 2002 to 2007 Hazard Ratio 2.94 (95% CI 1.87 to 4.62) C 2008 to 2013 Hazard Ratio 1.23 (95% CI 0.52 to 2.95) 32

34 Figure S21. Cumulative Incidence of Reoperation after Aortic-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 2.82 (95% CI 1.97 to 4.05) B 2002 to 2007 Hazard Ratio 2.30 (95% CI 1.54 to 3.42) C 2008 to 2013 Hazard Ratio 1.26 (95% CI 0.65 to 2.44) 33

35 Figure S22. Net Survival Penalty or Benefit of Bioprosthetic Mitral Valves Over Time (Additional Area Under the Survival Curve) A Years of Age B Years of Age C Years of Age Red circles represent point estimates of the difference in restricted mean survival time between bioprosthetic and mechanical valve recipients each year after surgery. Dashed lines are 95% confidence intervals. RMST, restricted mean survival time 34

36 Figure S23. Mortality After Mitral-Valve Replacement with Mechanical or Biological Prostheses Stratified by Decade of Age A Years of Age Hazard Ratio 1.18 (95% CI 0.95 to 1.48) B Years of Age Hazard Ratio 1.12 (95% CI 0.99 to 1.27) 35

37 Figure S24. Net Survival Penalty or Benefit of Bioprosthetic Mitral Valves Over Time (Additional Area Under the Survival Curve) in Patients and Years Old A Years of Age B Years of Age Red circles represent point estimates of the difference in restricted mean survival time between bioprosthetic and mechanical valve recipients each year after surgery. Dashed lines are 95% confidence intervals. RMST, restricted mean survival time 36

38 Figure S25. Mortality After Mitral-Valve Replacement with Mechanical or Biological Prostheses Stratified by Time Period Ages Years A 1996 to 2001 Hazard Ratio 1.21 (95% CI 1.01 to 1.45) B 2002 to 2007 Hazard Ratio 1.04 (95% CI 0.88 to 1.23) C 2008 to 2013 Hazard Ratio 0.81 (95% CI 0.62 to 1.08) 37

39 Figure S26. Cumulative Incidence of Stroke after Mitral-Valve Replacement, Stratified by Age Group and Type A Stroke in Patients Years Old Hazard Ratio 0.88 (95% CI 0.45 to 1.38) B Stroke in Patients Years Old Hazard Ratio 0.83 (95% CI 0.72 to 0.97) C Stroke in Patients Years Old Hazard Ratio 0.87 (95% CI 0.74 to 1.00) 38

40 Figure S27. Cumulative Incidence of Stroke after Mitral-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 0.67 (95% CI 0.48 to 0.94) B 2002 to 2007 Hazard Ratio 1.00 (95% CI 0.79 to 1.26) C 2008 to 2013 Hazard Ratio 0.97 (95% CI 0.72 to 1.30) 39

41 Figure S28. Cumulative Incidence of Bleeding after Mitral-Valve Replacement, Stratified by Age Group and Type A Bleeding in Patients Years Old Hazard Ratio 0.88 (95% CI 0.62 to 1.23) B Bleeding in Patients Years Old Hazard Ratio 0.80 (95% CI 0.72 to 0.90) C Bleeding in Patients Years Old Hazard Ratio 0.83 (95% CI 0.75 to 0.90) 40

42 Figure S29. Cumulative Incidence of Bleeding after Mitral-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 0.85 (95% CI 0.68 to 1.06) B 2002 to 2007 Hazard Ratio 0.72 (95% CI 0.61 to 0.85) C 2008 to 2013 Hazard Ratio 0.74 (95% CI 0.61 to 0.90) 41

43 Figure S30. Cumulative Incidence of Reoperation after Mitral-Valve Replacement, Stratified by Age Group and Type A Reoperation in Patients Years Old Hazard Ratio 3.35 (95% CI 1.76 to 5.92) B Reoperation in Patients Years Old Hazard Ratio 2.23 (95% CI 1.71 to 2.97) C Reoperation in Patients Years Old Hazard Ratio 1.51 (95% CI 1.09 to 2.10) 42

44 Figure S31. Cumulative Incidence of Reoperation after Mitral-Valve Replacement, Stratified by Time Period and Type Ages Years A 1996 to 2001 Hazard Ratio 2.58 (95% CI 1.78 to 3.74) B 2002 to 2007 Hazard Ratio 2.13 (95% CI 1.48 to 3.07) C 2008 to 2013 Hazard Ratio 1.35 (95% CI 0.78 to 2.34) 43

45 Table S1. Exclusion Criteria for Aortic-Valve Replacement Cohort Exclusion Category Concomitant or Prior Procedures Prior Procedures Diagnoses Non-ICD9 Characteristics ICD9-CM Code Procedure codes (from index or prior admissions) 335, 336, 3505, 3506, 3510, 3511, 3512, 3513, 3514, 3520, 3523, 3524, 3525, 3526, 3527, 3528, 3531, 3532, 3533, 3534, 3535, 3539, 3541, 3542, 3550, 3551, 3552, 3553, 3554, 3555, 3560, 3561, 3562, 3570, 3571, 3572, 3573, 3581, 3582, 3583, 3584, 3591, 3592, 3593, 3594, 3595, 3596, 3598, 3599, 3610, 3611, 3612, 3613, 3614, 3615, 3616, 3617, 3619, 3631, 3632, 3633, 3634, 3639, 3663, 3691, 3699, 3710, 3711, 3712, 3731, 3732, 3733, 3735, 3736, 3741, 3749, 3751, 3752, 3753, 3754,3755, 3760, 3762, 3763, 3764, 3765, 3766, 3767, 3834, 3835, 3844, 3845, 3973, 3979 Procedure codes (from prior admissions) 3520, 3521, 3522, 3523, 3524, 3525, 3526, 3527, 3528, 3531, 3532, 3533, 3534, 3535, 3539 Diagnosis codes (from index or prior admissions) 99602, 99603, 99671, 99683, V421, V422, V4321, V4322, V433, 41402, 41403, 41404, 41405, 41406, 41407, V4581, 7450, 74510, 74511, 74512, 74519, 7452, 7453, 7454, 7455, 74560, 74561, 74569, 7457, 7458, 7459, 74600, 74601, 74602, 74609, 7461, 7462, 7463, 7465, 7466, 7467, 74681, 74682, 74683, 74684, 74685, 74686, 74687, 74689, 7469, 7470, 74710, 74711, 74720, 74721, 74722, 74729, 74731, 74732, 74739, 74740, 74741, 74742, From index admission Age < 45, Age 65 Non-California State Resident (from patient zip code) Surgery after December 31, 2013 Dual coding for both valve types in same hospitalization More than 1 prosthesis implanted in the same day 44

46 Table S2. Exclusion Criteria for Mitral-Valve Replacement Cohort Exclusion Category Concomitant or Prior Procedures Prior Procedures Diagnoses Non-ICD9 Characteristics ICD9-CM Code Procedure codes (from index or prior admissions) 3521, 3522, 3525, 3526, 3527, 3528, 3510, 3511, 3512, 3513, 3531, 3532, 3533, 3534, 3535, 3539, 3834, 3835, 3844, 3845, 3973, 3979, 335, 336, 3712, 3751, 3752, 3753, 3754, 3755, 3760, 3762, 3763, 3764, 3765, 3766, 3767, 3731, 3732, 3735, 3710, 3711, 3691, 3699, 3741, 3749, 3631, 3632, 3633, 3634, 3639, 3550, 3551, 3552, 3553, 3554, 3555, 3560, 3561, 3562, 3663, 3570, 3571, 3572, 3573, 3541, 3542, 3581, 3582, 3583, 3584, 3591, 3592, 3593, 3594, 3595, 3596, 3598, 3599 Procedure codes (from prior admissions) 3520, 3521, 3522, 3523, 3524, 3525, 3526, 3527, 3528, 3531, 3532, 3533, 3534, 3535, 3539 Diagnosis codes (from index or prior admissions) 99602, 99671, 99683, V421, V422, V4321, V4322, V433, 41406, 41407, 41403, 41404, 41405, 99603, V4581, 7450, 74510, 74511, 74512, 74519, 7452, 7453, 7454, 7455, 74560, 74561, 74569, 7457, 7458, 7459, 74600, 74601, 74602, 74609, 7461, 7462, 7463, 7464, 7465, 7466, 7467, 74681, 74682, 74683, 74684, 74685, 74686, 74687, , 7470, 74710, 74711, 74720, 74721, 74722, 74729, 74731, 74732, 74739, 74740, 74741, 74742, From index admission Age < 40 years, Age 80 years Non-California State Resident (from patient zip code) Surgery before January 1, 1996 Surgery after December 31, 2013 Dual coding for both valve types in same hospitalization More than 1 prosthesis implanted in the same day 45

47 Table S3. Definitions of Baseline Characteristics* Baseline Comorbidity Hypertension Diabetes mellitus Coronary artery disease ICD9-CM Code Diagnosis codes (from index and prior admissions) 4011, 4019, 4020, 4021, 4029, 40290, 40291, 4030, 40300, 40301, 40310, 40311, 40390, 40391, 4040, 40400, 40401, 40402, 40403, 40410, 40411, 40412, 40413, 40490, 40491, 40492, 40493, 4050, 40509, 40511, 40519, 40591, 40599, and 4372 Diagnosis codes (from index and prior admissions) 2490, 24900, 24901, 24910, 24911, 24920, 24921, 24930, 24931, 24940, 24941, 24950, 24951, 24960, 24961, 24970, 24971, 24980, 24981, 24990, 24991, 25000, 25001, 25002, 25003, 25010, 25011, 25012, 25013, 25020, 25021, 25022, 25023, 25030, 25031, 25032, 25033, 25040, 25041, 25042, 25043, 25050, 25051, 25052, 25053, 25060, 25061, 25062, 25063, 25070, 25071, 25072, 25073, 25080, 25081, 25082, 25083, 25090, 25091, 25092, and Not revascularized Diagnosis codes (from index and prior admissions) 41000, 41001, 41002, 41010, 41011, 41012, 40120, 41021, 41022, 41030, 41031, 41032, 41040, 41041, 41042, 41050, 41051, 41052, 41060, 41061, 41062, 41070, 41071, 41072, 41080, 41081, 41082, 41090, 41091, 41092, 4110, 4111, 41181, 41189, 4130, 4131, 4139, 41400, 41401, 41402, 41403, 41404, 41405, 41406, 41407, and 4144 Prior percutaneous coronary intervention Diagnosis codes (from index and prior admissions) V4582 Procedure codes (from prior admissions) Peripheral vascular disease Cerebrovascular disease 0066, 1755, 3603, 3604, 3605, 3606, 3607, 3608, and 3609 Diagnosis codes (from index and prior admissions) 4400, 4401, 4402, 44020, 44021, 44022, 44023, 44024, 44029, 4403, 44030, 44031, 44032, 4404, 4408, 4409, and 4471 Diagnosis codes (from index and prior admissions) 36230, 36231, 36232, 36233, 36234, 36235, 36236, 36237, 34661, 34662, 34663, 43300, 43301, 43310, 43311, 4332, 43320, 43321, 4333, 43330, 43331, 43380, 43381, 43390, 43391, 43400, 43401, 43410, 43411, 43490, 43491, 4350, 46

48 Heart failure Atrial fibrillation/flutter Chronic obstructive pulmonary disease Chronic kidney disease 4351, 4352, 4353, 4358, 4359, 436, 437, 4370, 4371, 4372, 4373, 4374, 4375, 4376, 4377, 78043, 99702, 4378, 4379, 4380, 4381, 43810, 43811, 43812, 43813, 43814, 43819, 43820, 43821, 43822, 43830, 43831, 43832, 43840, 43841, 43842, 43850, 43851, 43852, 4386, 4387, 4388, 43881, 43882, 43883, 43884, 43885, 43889, and 4389 Diagnosis codes (from index and prior admissions) 39891, 4280, 4281, 42820, 42821, 42822, 42823, 42830, 42831, 42832, 42833, 42840, 42841, 42842, 42843, 40201, 40211, 40291, and 4289 Diagnosis codes (from index and prior admissions) 42731, Diagnosis codes (from index and prior admissions) 4910, 4911, 49120, 49121, 49122, 4918, 4919, 4920, 4928, 49300, 49301, 49302, 49310, 49311, 49312, 49320, 49321, 49322, 49381, 49382, 49390, 49391, 49392, 4940, 4941, and 496 Non-dialysis-dependent Diagnosis codes (from index and prior admissions) 40400, 40401, 40402, 40403, 40410, 40411, 40412, 40413, 40490, 40491, 40492, 40493, 5820, 5821, 5822, 5824, 5828, 58281, 58289, 5829, 5830, 5831, 5832, 5834, 5836, 5837, 5838, 58381, 58389, 5839, 5851, 5852, 5853, 5854, 5855, 5856, 5859, 586, 587, 5880, 5889, and V420 Dialysis-dependent Diagnosis codes (from index and prior admissions) Liver disease Cancer V560, V561, V562, V563, V5631, V5632, V568, V4511, V4512 Diagnosis codes (from index and prior admissions) 700, 701, 7020, 7021, 7022, 7023, 7030, 7031, 7032, 7033, 7041, 7042, 7043, 7044, 7049, 7051, 7052, 7053, 7054, 7059, 706, 7070, 7071, 709, 4560, 4561, 4562, 45620, 45621, 4563, 4564, 4565, 4566, 4568, 570, 5710, 5711, 5712, 5713, 57140, 57141, 57142, 57149, 5715, 5716, 5718, 5719, 5720, 5721, 5722, 5723, 5724, 5728, 5730, 5731, 5732, 5733, 5734, 5735, 5738, 7824, 7891, 78959, 7904, 7948, and V427 Oropharyngeal cancers Diagnosis codes (from index and prior admissions) 1400, 1401, 1403, 1404, 1405, 1406, 1408, 1409, 1410, 1411, 1412, 1413, 1414, 1415, 1416, 1418, 1419, 1420, 1421, 1422, 1428, 1429, 1430, 1431, 1438, 1439, 1440, 47

49 1441, 1448, 1449, 1450, 1451, 1452, 1453, 1454, 1455, 1456, 1458, 1459, 1460, 1461, 1462, 1463, 1464, 1465, 1466, 1467, 1468, 1469, 1470, 1471, 1472, 1473, 1478, 1479, 1480, 1481, 1482, 1483, 1488, 1489, 1490, 1491, 1498, and 1499 Gastrointestinal cancers Diagnosis codes (from index and prior admissions) 1500, 1501, 1502, 1503, 1504, 1505, 1506, 1508, 1509, 1510, 1511, 15012, 1513, 1514, 1515, 1516, 1518, 1519, 1520, 1521, 1522, 1523, 1528, 1529, 1530, 1531, 1532, 1533, 1534, 1535, 1536, 1537, 1538, 1539, 1540, 1541, 1542, 1543, 1548, 1550, 1551, 1552, 1560, 1561, 1562, 1568, 1569, 1570, 1571, 1572, 1573, 1574, 1578, 1579, 1580, 1588, 1589, 1590, 1591, 1598, and 1599 Respiratory cancers Diagnosis codes (from index and prior admissions) 1600, 1601, 1602, 1603, 1604, 1605, 1608, 1609, 161, 1610, 1611, 1612, 1613, 1618, 1619, 162, 1620, 1622, 1623, 1624, 1625, 1628, 1629, 163, 1630, 1631, 1638, 1639, 164, 1640, 1461, 1462, 1463, 1468, 1649, 165, 1650, 1658, and 1659 Bone and connective tissue cancer Diagnosis codes (from index and prior admissions) 1700, 1701, 1702, 1703, 1704, 1705, 1706, 1707, 1708, 1709, 1710, 1712, 1713, 1714, 1715, 1716, 1717, 1718, 1719, 1720, 1721, 1722, 1723, 1724, 1725, 1726, 1727, 1728, 1729, 17300, 17301, 17302, 17309, 17310, 17311, 17312, 17319, 17320, 17321, 17322, 17329, 17330, 17331, 17332, 17339, 17340, 17341, 17342, 17349, 17350, 17351, 17352, 17359, 17360, 17361, 17362, 17369, 17370, 17371, 17372, 17379, 17380, 17381, 17382, 17389, 17390, 17391, 17392, 17399, 1740, 1741, 1742, 1743, 1744, 1745, 1746, 1748, 1749, 1750, 1759, 1760, 1761, 1762, 1763, 1764, 1765, 1768, and 1769 Genitourinary cancers Diagnosis codes (from index and prior admissions) 179, 1800, 1801, 1808, 1809, 181, 1820, 1821, 1828, 1830, 1832, 1833, 1834, 1835, 1838, 1839, 1840, 1841, 1842, 1843, 1844, 1848, 1849, 185, 1860, 1869, 1871, 1872, 1873, 1874, 1875, 1876, 1877, 1878, 1879, 1880, 1881, 1882, 1883, 1884, 1885, 1886, 1887, 1888, 1889, 1890, 1891, 1892, 1893, 1894, 1898, and 1899 Lymphoid cancers 48

50 Diagnosis codes (from index and prior admissions) 20000, 20001, 20002, 20003, 20004, 20005, 20006, 20007, 20008, 2001, 20010, 20011, 20012, 20013, 20014, 20015, 20016, 20017, 20018, 2002, 20020, 20021, 20022, 20023, 20024, 20025, 20026, 20027, 20028, 20030, 20031, 20032, 20033, 20034, 20035, 20036, 20037, 20038, 20040, 20041, 20042, 20043, 20044, 20045, 20046, 20047, 20048, 20050, 20051, 20052, 20053, 20054, 20055, 20056, 20057, 20058, 20060, 20061, 20062, 20063, 20064, 20065, 20066, 20067, 20068, 20070, 20071, 20072, 20073, 20074, 20075, 20076, 20076, 20077, 20078, 20080, 20081, 20082, 20083, 20084, 20085, 20086, 20087, 20088, 20100, 20101, 20102, 20103, 20104, 20105, 20106, 20107, 20108, 20110, 20111, 20112, 20113, 20114, 20115, 20116, 20117, 20118, 20120, 20121, 20122, 20123, 20124, 20125, 20126, 20127, 20128, 20140, 20141, 20142, 20143, 20144, 20145, 20146, 20147, 20148, 20150, 20151, 20152, 20153, 20154, 20155, 20156, 20157, 20158, 20160, 20161, 20162, 20163, 20164, 20165, 20166, 20167, 20168, 20170, 20171, 20172, 20173, 20174, 20175, 20176, 20177, 20178, 20190, 20191, 20192, 20193, 20194, 20195, 20196, 20197, 20198, 20200, 20201, 20202, 20203, 20204, 20205, 20206, 20207, 20208, 20210, 20211, 20212, 20213, 20214, 20215, 20216, 20217, 20218, 20220, 20221, 20222, 20223, 20224, 20225, 20226, 20227, 20228, 20230, 20231, 20232, 20233, 20234, 20235, 20236, 20237, 20238, 20240, 20241, 20242, 20243, 20244, 20245, 20246, 20247, 20248, 20250, 20251, 20252, 20253, 20254, 20255, 20256, 20257, 20258, 20260, 20261, 20262, 20263, 20264, 20265, 20266, 20267, 20268, 20270, 20271, 20272, 20273, 20274, 20275, 20276, 20277, 20278, 20280, 20281, 20282, 20283, 20284, 20285, 20286, 20287, 20288, 20290, 20291, 20292, 20293, 20294, 20295, 20296, 20297, 20298, 20300, 20301, 20302, 20310, 20311, 20312, 2038, 20380, 20381, and Hematologic cancers Diagnosis codes (from index and prior admissions) 20400, 20401, 20402, 20410, 20411, 20412, 20420, 20421, 20422, 20480, 20481, 20482, 20490, 20491, 20492, 20500, 20501, 20502, 20510, 20511, 20512, 20520, 20521, 20522, 20530, 20531, 20532, 20580, 20581, 20582, 20590, 20591, 20592, 20600, 20601, 20602, 20610, 20611, 20612, 20620, 20621, 20622, 20680, 20681, 20682, 20690, 20691, 20692, 20700, 20701, 20702, 20710, 20711, 20712, 20720, 20721, 20722, 20780, 20781, 20782, 20800, 20801, 20802, 20810, 20811, 20812, 2082, 20820, 20821, 20822, 20880, 20881, 20882, 20890, 20891, 20892, 20900, 20901, 20902, 20903, 20910, 20911, 20912, 20913, 20914, 20915, 20916, 20917, 20920, 20921, 20922, 20923, 20924, 20925, 20926, 20927, 49

51 20929, 20930, 20931, 20932, 20933, 20934, 20935, 20936, 20940, 20941, 20942, 20943, 20950, 20951, 20952, 20953, 20954, 20955, 20956, 20957, 20960, 20961, 20962, 20963, 20964, 20965, 20966, 20967, 20969, 20970, 20971, 20972, 20973, 20974, 20975, and Other cancers Diagnosis codes (from index and prior admissions) Osteoporosis Hip fracture 1900, 1901, 1902, 1903, 1904, 1905, 1906, 1907, 1908, 1909, 191, 1910, 1911, 1912, 1913, 1914, 1915, 1916, 1917, 1918, 1919, 1920, 1921, 1922, 1923, 1928, 1929, 193, 1940, 1941, 1943, 1944, 1945, 1946, 1948, 1950, 1951, 1952, 1953, 1954, 1955, 1958, 1960, 1961, 1962, 1963, 1965, 1966, 1968, 1969, 1970, 1971, 1972, 1973, 1974, 1975, 1976, 1977, 1978, 1980, 1981, 1982, 1983, 1984, 1985, 1986, 1987, 19881, 19882, 19889, 199, 1990, 1991, and 1992 Diagnosis codes (from index or prior admissions) 7330, 73300, 73301, 73302, 73303, Fractures of the femur Diagnosis codes (from index or prior admissions) 82000, 82001, 82002, 82003, 82009, 82010, 82011, 82012, 82013, 82019, 82020, 82021, 82022, 82030, 82031, 82032, 8208, 8209 Fractures of the pelvis Diagnosis codes (from index or prior admissions) 8080, 8081, 8082, 8083, 80841, 80842, 80843, 80844, 80849, 80851, 80852, 80853, 80854, 80859, 8088, 8089 Malnutrition Anemia Prior myocardial infarction Diagnosis codes (from index or prior admissions) 260, 261, 262, 2630, 2631, 2632, 2638, 2639, 2699 Diagnosis codes (from index or prior admissions) 2800, 2801, 2808, 2809, 2810, 2811, 2812, 2813, 2814, 2818, 2819, 2822, 2823, 2828, 2829, 2830, 28310, 28319, 2839, 28409, 28489, 2849, 2850, 2851, 28521, 28522, 28529, 2853, 2858, 2859 Diagnosis codes (from prior admissions) 41001, 41011, 41021, 41031, 41041, 41051, 41081, , 41012, 41022, 41032, 41042, 41052, 41082, , 41010, 41020, 41030, 41040, 41050, 41080,

52 Hypothyroidism Ever Smoker Obesity Asthma Dementia Aortic stenosis Aortic insufficiency Aortic valve disorder Mitral stenosis Mitral regurgitation Mitral stenosis and regurgitation Mitral valve disorder Tricuspid valve disease Diagnosis codes (from index admission) 41002, 41012, 41022, 41032, 41042, 41052, 41082, Diagnosis codes (from index or prior admissions) 244, 2440, 2441, 2442, 2443, 2448, 2449 Diagnosis codes (from index or prior admissions) 305.1, V15.82 Diagnosis codes (from index or prior admissions) 278, 2780, 27800, 27801, 27802, Diagnosis codes (from index or prior admissions) 493, 4930, 49300, 49301, 49302, 4931, 49310, 49311, 49312, 49320, 49321, 49322, 4938, 49381, 49382, 4939, 49390, 49391, Diagnosis codes (from index or prior admissions) 2900, , 29011, 29012, 29013, 29020, 29021, 2903, 29040, 29041, 29042, 29043, 29410, 29411, 29420, 29421, 3310, 33119, 33182, 2912 Diagnosis codes (from index admission) 3950, 3960, 3962, 7463 Diagnosis codes (from index admission) 3951, 3961, 3963, 7464 Diagnosis codes (from index admission) 3952, 3968, 3969, 4241 Diagnosis codes (from index admission) 3940, 3960, 3961, 7465 Diagnosis codes (from index admission) 3941, 3962, 3963, 7466 Diagnosis codes (from index admission) 3942 Diagnosis codes (from index admission) 3949, 3968, 3969, 4240 Diagnosis codes (from index admission) 3970 *Modified from Chikwe et al. JAMA