An Example of Business Analytics in Healthcare Colleen McGahan Biostatistical Lead Cancer Surveillance & Outcomes BC Cancer Agency cmcgahan@bccancer.bc.ca
Improve Ovarian Cancer Outcomes Business relevancy Analytics Actionable insight Performance measurement Delivers value
An Example of Business Analytics in Healthcare Overview of Ovarian Cancer Background to Business Relevancy Analytics Conclusions
OVARIAN CANCER OVERVIEW
Ovarian Cancer Is not as common as other cancers Roughly 300 cases of ovarian cancer are diagnosed each year compared to 3,000 cases of breast cancer It is more aggressive than other cancers 1-year survival is just over 75%, and given a women survived the 1 st year, their 5-year survival is 47% Compared to 97% and 86%, respectively for breast cancer
Ovarian Cancer Surgical removal of all cancer can be difficult Think of grains of sand being spread throughout the abdominal cavity as well as multiple masses
Ovarian Cancer
Ovarian Cancer To increase survival all the cancer has to be removed (optimal debulking) Sometimes it s just not possible, so masses should be removed to within 1cm in size Any tumour left which is larger than 1cm is thought to be sub-optimal surgery
Ovarian Cancer Stage of disease is determined by how far the cancer has spread Grade of disease is based on the severity of the cells unable to differentiate normal cells gives a higher grade Histology is the type of cell involved
Ovarian Cancer Surgery can be done by: General Surgeon They do a large variety of surgeries, not always cancer surgery OR Gynaecologic Oncology Surgeon Specialised in cancer surgery for gynaecology (ovarian, uterus, cervical etc)
Ovarian Cancer Surgery is almost always followed by chemotherapy Combination chemotherapy (rather than single agent) is considered the standard of care.
BACKGROUND TO BUSINESS RELEVANCY
BC Cancer Agency Mandate To provide a province-wide, populationbased cancer control program for BC
BC Cancer Agency Mission To reduce the incidence of cancer To reduce the mortality rate of people with cancer To improve the quality of life of people living with cancer
Cancer Surveillance & Outcomes Perform cancer control statistical analyses and reporting for: surveillance business planning performance measurement
Some statistics produced Incidence Mortality Survival Projections of incidence/mortality ------------------------------------------------------------- By: Gender Age Cancer disease site [eg. Breast, prostate etc] Regional area [eg. Health Authority (HA) etc] ---------------------------------------------------------------------- and many more
Relative Hazard Ratio Statistic that prompted this study 1.75 Hazard Ratio for Death due to Ovarian Cancer Relative to the Province for Women Diagnosed with Ovarian Cancer over a 3-year Period 1.5 1.25 1 0.75 0.5 0.25 0 HA #1 HA #2 HA #3 HA #4 HA #5 Health Authority
Why are the outcomes for women in the one health authority different to all the others? What do we need to do to change it? Need to drill down further for more information
Data elements extracted Patient demographics Age, residence location at time of diagnosis Disease information Stage, grade, histology type Treatment Type of chemotherapy, surgery, optimal debulking Physician Type who assessed the patient, type who performed the surgery Outcome Overall survival
Analysis In order to do this, we want to understand what factors are influencing the patient s overall survival. Used: PROC LIFETEST to look at Kaplan Meier Survivor Functions by strata PROC PHREG to perform Cox Regression Utilised the BASELINE statement in PROC PHREG to obtain predicted survival functions
Findings Four key factors were identified to be associated with lower overall survival: Advanced stage disease (stage 3C/4 v 1/2/3A/3B) Type of chemo (single v combination) Grade (grade 3 v 1/2) Extent of Surgery (sub-optimal v optimal debulking)
Overall Survival (OS) by Stage Low Stage Disease Advanced Stage Disease Survival (Months)
OS: HA#4 compared to HA's 1,2,3 & 5, by Stage of Disease Median OS (Mths) for Stage 3C/4: HA #1,2,3,5 = 30mths (26, 35) Low Stage: Green HA#4 Blue HA #1,2,3,5 HA #4 = 21mths (16, 29) High Stage: Brown - HA #1,2,3,5 Olive HA#4 Survival (Months)
High Stage disease by HA HA region Stage 3C/4 1 69.4% 2 55.9% 3 50.3% 4 58.9% 5 64.3% (58.6%) (p=0.0048)
Findings Four key factors were identified to be associated with lower overall survival: Advanced stage disease (stage 3C/4 v 1/2/3A/3B) Type of chemo (single v combination) Grade (grade 3 v 1/2) Extent of Surgery (sub-optimal v optimal debulking)
Overall Survival by Type of Chemotherapy Red = No Chemo Blue = Combination Chemo Green = Single Agent Chemo Survival (Months)
OS by Type of Chemotherapy for Low Stage Red = No Chemo Blue = Combination Chemo Green = Single Agent Chemo Survival (Months)
OS by Type of Chemotherapy for High Stage Red = No Chemo Blue = Combination Chemo Green = Single Agent Chemo Survival (Months)
Chemotherapy for Advanced Stage Disease (3C/4) HA region None Single Combination 1 10.2% 15.8% 74.0% 2 7.3% 11.3% 81.4% 3 8.0% 12.5% 79.5% 4 13.4% 31.1% 55.5% 5 5.6% 5.6% 88.9% P=0.0002
Chemotherapy for Advanced Stage Disease (3C/4) HA region None Single Combination 1 10.2% 15.8% 74.0% 2 7.3% 11.3% 81.4% 3 8.0% 12.5% 79.5% 4 13.4% 31.1% 55.5% 5 5.6% 5.6% 88.9% P=0.0002
Findings Four key factors were identified to be associated with lower overall survival: Advanced stage disease (stage 3C/4 v 1/2/3A/3B) Type of chemo (single v combination) Grade (grade 3 v 1/2) Extent of Surgery (sub-optimal v optimal debulking)
OS by Grade Blue = Low Grade Red = High Grade Survival (Months)
OS: HA#4 compared to HA s 1,2,3 & 5 for Low Grade Disease Low Grade Disease: Blue HA #1,2,3,5 Red HA#4 Survival (Months) Survival (Months)
OS: HA#4 compared to HA s 1,2,3 & 5 for High Grade Disease High Grade Disease: Blue HA #1,2,3,5 Red HA#4 Survival (Months) Survival (Months)
Grade of Disease HA region Grade 3 1 75.8% 2 76.5% 3 77.4% 4 75.4% 5 87.5% 76.6% (p=0.7612)
Findings Four key factors were identified to be associated with lower overall survival: Advanced stage disease (stage 3C/4 v 1/2/3A/3B) Type of chemo (single v combination) Grade (grade 3 v 1/2) Extent of Surgery (sub-optimal v optimal debulking)
OS by Extent of Surgery Green = Optimal Debulk Olive = Optimal Debulk to within 1cm Blue = Sub-optimal Surgery Red = No Surgery Survival (Months)
OS: HA#4 compared to HA s 1,2,3 & 5 Optimal Debulk Blue = All others Red = HA#4 Survival (Months)
OS: HA#4 compared to HA s 1,2,3 & 5 Optimally Debulked to within 1cm Blue = All others Red = HA#4 Survival (Months)
OS: HA#4 compared to HA s 1,2,3 & 5 Sub-optimal Surgery Blue = All others Red = HA#4 Survival (Months)
OS: HA#4 compared to HA s 1,2,3 & 5 No Surgery Survival (Months) Blue = All others Red = HA#4
Sub-Optimal Debulk by Stage of Disease HA region Low Stage Advanced Stage 1 2% 43.0% 2 2% 38.3% 3 2% 40.6% 4 5% 60.2% 5 10% 21.4% P=0.4143 P=0.0042
Sub-Optimal Debulk by Stage of Disease HA region Low Stage Advanced Stage 1 2% 43.0% 2 2% 38.3% 3 2% 40.6% 4 5% 60.2% 5 10% 21.4% P=0.4143 P=0.0042
Why are the outcomes for women in the one health authority different to all the others? What do we need to do to change it?
Conclusions The outcomes are worse for patients with advanced stage disease within HA#4 compared to the rest of the province. Within advanced stage disease patients in HA#4: a lower proportion of patients are receiving the standard of care combination chemotherapy. there is a higher rate of sub-optimal surgery.
Survival Probability Predicted Overall Survival had Patients in HA#4 been given Standard of Care 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 All others HA#4 - Predicted OS HA#4 - Observed OS 0.2 0.1 0 0 10 20 30 40 50 60 70 Survival (Months)
Actionable Insight A Gyne Oncology Retreat was held included; surgeons, radiation oncologists and medical oncologists new gynecologic oncology surgeon and a medical oncologist have started A new surgeon and medical oncologist are now working in HA #4
Any Questions? Colleen McGahan cmcgahan@bccancer.bc.ca
PROC LIFETEST ods graphics on; title 'Overall Survival by Chemo'; proc lifetest data=anal.analysis plots=(survival (test) lls) timelist=(12 to 60 by 12); time surv*surv_cens(1); strata chemo; run; ods graphics off; Requests the plots that you want Survival curve & log ve log of estimated survival function to help check for proportionality Requests the p-value of the homogeneity test specified in the strata statement to be displayed on plot
PROC PHREG proc phreg data=ovary; where gynonc ne 9 and new_optdebulk ne. and gynonc_assess ne 9; format figo $stg3i. grade grade. new_optdebulk debulk_n. gynonc_assess asses2i. gynonc gynonc2i. chemo chemo. histgp $hist2gp.; class figo(descending) grade(descending) new_optdebulk schemo(descending) nchemo(descending) histgp; model surv*surv_cens(1) = figo grade new_optdebulk schemo nchemo schemot nchemot new_optdebulk*grade / rl; schemot=schemo*log(surv); nchemot=nchemo*log(surv); hazardratio new_optdebulk / at (grade=all); run;
Surgery HA region Assess by Gyn Onc Surgeon type Gen Gyn/ Gen Surg Gyn Onc No surgery 1 (58.5%) 34.2% 43.1% 22.7% 2 (70.3%) 29.4% 58.7% 11.9% 3 (79.9%) 15.0% 71.1% 13.9% 4 (67.3%) 28.2% 61.4% 10.4% 5 (75%) 28.6% 57.1% 14.3% Total (69.2%) (p=0.0003) P=0.0001
Chemotherapy for high-risk early stage and advanced stage HA region None Single Combination 1 17.0% 13.0% 70.0% 2 9.8% 8.2% 82.0% 3 10.7% 7.7% 81.6% 4 15.1% 23.4% 61.5% 5 14.8% 3.7% 81.5% P<.0001
Cox Regression Model Variable Parameter HR [95% CI] Stage Stage IV 2.13 [1.37, 3.31] Stage IIIC 1.67 [1.18, 2.44] Chemo Single vs. combination 5.17 [2.62, 12.06] None vs. single agent 3.03 [1.50, 6.10] None vs. combination 17.01 [8.00, 36.18] Grade 3 0 residual vs. 1 cm 0.63 [0.43, 0.93] Subopt vs. 1 cm 1.51 [1.11, 2.06] Subopt vs. no surgery 0.46 [0.34, 0.62]
Prognostic Factors for Overall Survival in Ovarian Cancer Age Grade [1, 2 or 3; Gr 3 is high risk) Stage [1A, 1B, 1C, 2, 3A, 3B (low risk); 3C, 4 (high risk)] Histotype [non-serous, serous (high risk)] Surgeon [General, Gynecologic Oncologist] Optimal debulking (no surgery, sub-optimal, within 1cm, no residual) Chemotherapy (none, single, combination)
Disease characteristics HA region Median age Grade 3 Histology Serous Stage 3C/4 1 65.0 (75.8%) (75.3%) (69.4%) 2 60.0 (76.5%) (66.9%) (55.9%) 3 58.0 (77.4%) (62.9%) (50.3%) 4 65.0 (75.4%) (71.5%) (58.9%) 5 59.5 (87.5%) (82.1%) (64.3%) 62.0 years (p=0.0002) (76.6%) (p=0.7612) (69.4%) (p=0.0434) (58.6%) (p=0.0048)
Extent of Debulking in Advanced Stage HA region Suboptimal debulk 1 43.0% 2 38.3% 3 40.6% 4 60.2% 5 21.4% P=0.0036
Ovarian Cancer (1) Cancer of the right ovary (2) Metastases on the intestines (3) Metastases against the diaphragm (4) Metastases on the omentum ( apron of fatty tissue which is attached to the stomach and which lies over the small bowel [NEED TO SPECIFY DIAGRAM REFERENCE]