Author's response to reviews

Transcription

1 Author's response to reviews Title: Patterns of ambulatory medical care utilization in elderly patients with special reference to chronic diseases and multimorbidity - Results from a claims data based observational study in Germany Authors: Hendrik van den Bussche (bussche@uke.de) Gerhard Schön (g.schoen@uke.de) Tina Kolonko (tina.kolonko@gmx.de) Heike Hansen (h.hansen@uke.de) Karl Wegscheider (k.wegscheider@uke.de) Gerd Glaeske (gglaeske@zes.uni-bremen.de) Daniela Koller (dkoller@zes.uni-bremen.de) Version: 3 Date: 6 July 2011 Author's response to reviews: see over

2 Cover Letter MS: Patterns of ambulatory medical care utilization in elderly patients with special reference to chronic diseases and multimorbidity - Results from a claims data based observational study in Germany van den Bussche Hendrik, Schön Gerhard, Kolonko Tina, Hansen Heike, Wegscheider Karl, Glaeske Gerd, Koller Daniela Corresponding author: Hendrik van den Bussche Adress of corresponding author: Institute of Primary Medical Care University Medical Center Hamburg-Eppendorf Martinistrasse 52 D Hamburg Germany Tel Mobile: Fax bussche@uke.de Dear editor, On behalf of the co-authors, I submit the revised manuscript of the above paper based on the most valued comments of the reviewers. The revision profited very much from them. The paper has been overdone almost completely. The answers to the comments and proposals of the reviewers are listed in the following table. In the paper itself, the major changes related to the comments of reviewer 1 are in turquoise, and those related to reviewer 2 in yellow. The manuscript has also been revised by a native speaker. Sincerely, Hendrik van den Bussche 1

3 Row Nr. Reviewer 1 comments Author s response What is the relationship between disability and degree of comorbidity (number of chronic conditions)? Are they really independent variables or is there coufounding? Assuming nonindependence, how was this handled in the regression analyses? Please make this clear Requiring diagnoses to be present in three quarters meant that people with fewer visits were excluded from being considered to have multimorbidity. Would this not artifically increase the number of visits in those with mm versus nmm? The fact that specialist internists were included as primary care physicians certainly undercounts people who saw specialists. Is there anyway to account for this serious limitation? (At least it has to be recognized) What percentage of PCPs are family physicians versus internists? What proportion of visits are for prescriptions? Doesn't this have to be considered in the analysis and intepretation? These variables are very probably non-independent. The regression analysis, however, allows to measure the influence of one variable while all others are kept constant. In such a way, confounding can be controlled for. Therefore, the influence of the number of chronic conditions and the presence of nursing dependency (as a proxi for ADL-related disability) is correctly expressed by the regression coefficients for both variables. When people would see a physician with three or more of the chronic conditions during 1 or 2 quarters of the year instead of 3, they would be classified as non-multimorbid in this study. As there is no unique definition of multimorbidity, every yes/no criterion setting provokes artificial assignments of individuals. A code for minimum 3 chronic diseases in at least 3 of 4 quarters within the one-year observation period was chosen to minimize erroneous or transitory diagnoses by physicians and thus increase validity. This selection criterion is almost standard in all claims data based studies in Germany (see Swart & Ihle. Routinedaten im Gesundheitswesen. Handbuch Sekundäranalyse. Huber Bern 2005). The criterion of three chronic conditions was considered to be a more valid cut-off score for multimorbidity in elderly patients treated in the ambulatory care setting instead of the usual two chronic conditions criterion, as the former frequently leads to very high rates of multimorbidity in the age group over 65 years (also in our data set). Recent research supports using the criterion of 3 conditions for investigations of multimorbidity in the ambulatory care setting, especially when the aim is to compare the multimorbid sample and the nonmultimorbid sample [Schram MT et al.: Setting and registry characteristics affect the prevalence and nature of multimorbidity in the elderly. J Clin Epidemiol 2008, 61: Fortin M et al.: Prevalence estimates of multimorbidity: a comparative study of two sources. BMC Health Serv Res 2010, 10: 111). The question of the selection criterion is described in detail in the methods section. This database problem does in fact not allow to discern between referrals between the internists working as PCPs and those working as specialized internists. Therefore, this limitation does not allow to analyse the referral processes related to multimorbidity as far as the role of single disciplines is concerned, unless complicated analyses, estimations and interpretations are introduced. Furthermore, a very detailed knowledge of the rules concerning referrals in the German statutory insurance system in necessary in order to interpret these data. For al these reasons, we decided to exclude this topic from this publication and to dedicate a special publication (in German) to this complicated matter. At January 1, 2010, GPs and general internists acted as PCPs, corresponding to a relation of 3.6:1. The data do not allow to differentiate between visits for prescriptions and visits for other reasons. This could only be clarified by detailed observational studies as it is the physician who decides if and eventually how long he/she will communicate with the patient before renewing the prescription. This might vary between no contact, hallo -contact, short interview and longer consultation including clinical examination. The problem described by the reviewer is a standard problem in nearly all (comparative) studies on consultation frequency in medical care based on administrative data. The point that a (unknown) proportion of the contacts reported in this paper may have been for renewal of prescriptions only has been added in the discussion. The point is taken to the methods section. 2

4 6 7 Standard deviations should be included in the figures. Recent studies from Canada and Israel are showing that it is not the number of chronic conditions that account for high costs but, rather, the morbidity mix, as measured by the ACG casemix measure. (..) In this regard it is noteworthy that different triads in your study are associated with different numbers of consultations: the ones with DIFFERENT types of conditions are associated with more visits than the triads with similar and probably related conditions. That is, it is likely that it is the variety of conditions (rather than their number that is associated with high use. This is an issue that deserves recognition and perhaps advocating use of morbidity mix methods rather than counts of conditions. Included as proposed. Also, in the additional files, confidence intervals have been included in the graphs on the number of contacts and of contacted physicians according to the individual chronic conditions in oder to give a more clear picture of statistical significance of diffderences. The opinion of the reviewer that the triads with different types of conditions are associated with more visits than the triads with similar and probably related conditions is not very well supported by the data. We have therefore ranked the contact frequency for the 50 most prevalent triads according to numbers of contact in additional file 4. Looking at the six most frequent triads for example, the view of the reviewer is confirmed by triads 49 (partly), 18 and 20 but not by the immediate followers 31, 44 and 48. The question of reasons for differences in contact frequency between individual patterns has been taken up in the discussion section. 3

5 8 Reviewer 2 comments 1. Discussion The results are interesting, but the discussion needs further improvement. Not all aims and (important) results were addressed in the discussion, some topics were discussed in several paragraphs and not combined in one paragraph (e.g. frequency of contacts currently discussed in paragraph 2 & 8) and some paragraphs were unclear (paragraph 11). I suggest the following scheme: Paragraph one: - very brief summary of the aims - What was the frequency of contacts and number of physicians contacted - What was the influence of the other factors (age, sex, number of multimorbidity pattern and nurse dependency) on utilization - What was the association between utilization of primary care physicians and specialists Paragraph two: - Elaborate on the frequency of contacts and number of physicians contacted - In comparison with other countries Paragraph three: the influence of other factors on utilization - Age - Sex - Number of chronic diseases - Multimorbidity pattern - Nurse dependency Paragraph four: association between utilization of primary care physicians and specialists in comparison with other countries Paragraph five: weaknesses and strengths of this study Paragraph six: conclusions Author s response The discussion section has been completely overdone based on the suggestions of the reviewer: Paragraph one: - very brief summary of the aims - What was the frequency of contacts and number of physicians contacted - What was the influence of the other factors (age, sex, number of multimorbidity pattern and nurse dependency) on utilization. Paragraph two: the influence of other factors on utilization: - Age - Sex - Number of chronic diseases - Multimorbidity patterns - Nurse dependency Paragraph three: Frequency of contacts and number of physicians contacted in comparison with other countries Paragraph four: weaknesses and strengths of this study Paragraph five: conclusions 4

6 Not all aims correspond clearly with the methods, reported results and the discussion: In the aims it is mentioned that also the utilization of primary care physicians in relation to specialists is investigated. - In the methods it is not explained how this is investigated - In the results you discover it is investigated via referral rates, although the subtitle size, mode and target of referral is not helpful. I suggest changing this subtitle. - In the discussion it is concluded in paragraph 7 that the results indicate a centering of utilization. However, the association between utilization of primary care physicians and specialist is only discussed in paragraph 9. The analysis of referrals. Was omitted (see heading 3) Not all aims correspond clearly with the methods, reported results and the discussion: Another aim that is mentioned in the abstract (and not in the introduction) is that the ambulatory medical care utilization will be investigated in relation to patterns of mulitmorbidity. - In the methods it is not explained how these patterns of mulitmorbidity are investigated (e.g. it is not explained how the triadic combinations were calculated) - In the discussion multimorbidity patterns are not discussed. However, in the results section (even in the results presented in the abstract) numbers for contact frequencies are also shown for triadic combinations. - Strikingly, the number of contacts per year was lower for triadic combinations than for certain separate chronic diseases, also this was not commented on in the discussion (abstract results line 5-7) The data is seven years old. In the methods it is mentioned that the authors had access to the GEK data from 2006 (paragraph 2 line 10). Why wasn t more recent data used? 3.2. Paragraph 1 line 1: The term cohort is used, but on the other hand it is also stated that a cross-sectional approach was used (Discussion paragraph 4 line 11 13). Hence, the term cohort is not appropriate and the crosssectional approach of this study should be mentioned in the methods In the methods section it is now explained how the patterns of mulitmorbidity are investigated and how the triadic combinations were identified. Similarities between the results for triads and singe conditions are explained in the methods section. In the abstract, this similarity was stressed. The point of utilization in triadic combinations was taken up in the discussion section. The dataset covers the period , delivered to us in several consecutive mails by GEK. The study of 2004 is in our view a baseline study which will be followed by longitudinal studies. The fact that it took a long period of time to finish the first study is due to the fact that it was very complicated to assemble the different subsets of each year (subsets for diagnoses, physician contacts, hospital contacts, drugs etc.). Choosing the year 2006 for defining the diagnoses to be included was due to the fact that the other sources used in this process used data from 2006 and therefore we selected the comparable GEK-Dataset. Our statistics department considers this study as a cohort study since the observation period is 12 months. However, we do not measure at several time points but we aggregate the number of contacts and other parameters for the whole 12 months period. Of course, it was not logical to use the term cohort and speak about a cross-sectional approach in the discussion section in the meantime. What was meant was that we would need a several years observation period in order to investigate the possibility of healthy survival. As a consequence: In order to avoid further discussions on terminology the term study population was introduced instead of cohort. This new term is perfectly suitable for this presentation. The period under investigated was included in the methods section. The text on the healthy survivor problem was rewritten in the discussion section. 5

7 No explanation of the multimorbidity and non-multimorbidity sample: In the abstract and results the term sample is used for patients with and without mulitmorbidity, this suggest that not all patients with and without mulitmorbidity were selected. However, in the methods it is not explained how the multimorbidity sample and non-multimorbidity sample was created 3.4. Missing explanation of how multimorbidity patterns were studied/how the triadic combinations were created. The way of creation of the multimorbidity and the non-multimorbidity sample is explained in the methods section. The population under study was diveded into the mm sample and the nmm sample, which implies that all patients are included, either in one or in the other sample. Now explained in the methods section Missing explanation of how (self-) referral rates were calculated. Section on referrals omitted (see heading 3) 4. Results The term subsample has been replaced by (mm or nmm) sample in the whole subsamples have not been explained in the methods. paper 1. Results 1.1. subtitle 2: Frequency of contact with physicians paragraph 1 Table 1 lists the number of contacts for the nmm- and the mm-sample according to sex. The translation into weeks/months was included in the text. Line 9: 17 I would love to see a table for the different age groups (subdivided for total cohort, mm-sample, nmm-sample, but not necessarily) stating the percentages of different contact frequencies in terms of contact per week or month Abstract 1.1. Aims: Please include some background that justifies this study (e.g. Background paragraph 6 line 1: in order to estimate the future demands for health services, the analysis of current utilization patterns of the elderly and their determinants is crucial. Therefore the aim of this study is ) 1 Abstract 1.2. Methods: the authors should choose either to include more details in the methods or to simplify the presented results (of the abstract). For instance, the term multimorbidity needs some explanation? Also multivariate methods can t this be replaced by multivariable linear regression? 1 Abstract 1.3. Results: the total cohort and the multimorbidity sample are used in the results, but not explained in the methods. For the understanding of the results (in the abstract) it is not necessary to mention the total cohort or the mm-sample (e.g. this proportion was higher among patients with less than 3 chronic diseases/without mulitmorbidity). Included as suggested Included as suggested Overdone as suggested 6

8 2. Background corrected Paragraph 1 line 4 & 5 versus 6: overlap between these two sentences? 2.2. Paragraph 3 line 3: This sentence is not clear: For example, many studies on utilization related to age and gender are based on survey data 22 with a recall-biased a posteriori investigation of utilization. Aren t the last two words to be deleted? 2.3. Paragraph 5 line 4 & 5: I don t understand this sentence. Is fist 23 the correct word here? 3. Methods 3.1. Paragraph 1 line 8: previous studies have shown that results from 24 the GEK database can be transferred to the German population as a whole if age and gender adjusted. The references are missing Paragraph 3 line 4: spelling mistake: in Germany corrected Paragraph 6 line 1: in the text nurse dependency is also referred to as disability (e.g. background paragraph 6 line 4). If you prefer disability above nurse dependency than it should be explained in the methods that nurse dependency was included as a proxy for disability. 4. Results 4.1. Subtitle 1: Sociodemographic structure of the sample Line 7: women were overrepresented # I prefer there were more women than men with multimorbidity Present these results in a table so that you can decide for yourself whether you want to see the difference in number of diseases or sex in the mm-sample, the nmm-sample or the total cohort Subtitle 2: Frequency of contact with physicians Paragraph 1 Line 9: I would take advantage of the way of writing the frequency of contacts in terms of weeks or months. The results are more striking when presented as One contact every week or every two weeks Paragraph 2 line 3 5: difficult sentence, I had to read it a few times before I understood it Paragraph 2 line 7 8: I would delete (e.g. 5 contacts per year in the mm-sample) although the mean age difference between the youngest and the oldest age-groups is 17 years. It made it more difficult, while looking at the table is clear enough. Two words deleted as proposed Typing error: GP as the first physician in a quarter Reference introduced as suggested The relationship between disability and nursing dependency is explained in the methods section. Except for some comments in the discussion section, the results are related to nursing dependency only because disability cannot be derived directly from the insurance data. Sample replaced by study population Corrected as proposed See new table 1 Corrected as proposed Reworded as suggested Deleted as proposed; taken over to the discussion section 7

9 Paragraph 3 line 4: indexed shouldn t it be selected chronic conditions as indexed refers to the complete ICD index? 4.3. Subtitle 4: Size mode and target of referral The choice of the subtitle is strange as this was not mentioned in the aims/methods. Maybe the utilization of primary care physicians in relation to specialists is less confusing? 5. Discussion 5.1. Paragraph 1 line 3: 2/3 more physicians it was hard to interpret what it means that patients with mulitmorbidity have 2/3 more contact compared to patients without mulitmorbidity. Maybe try to formulate it differently. In this study, the term index is not related to the ICD-index. Here, an indexed condition is an individual chronic condition from the multimorbidity patterns selected for a specific analysis (e.g. index condition hypertension means that we investigated all persons with hypertension in their multimorbidity pattern). We now used the term individual as we did not select chronic conditions from the 46er list but reported the data for all of them Section on referrals omitted (se heading 3) Replaced by two third Paragraph 3 line 1: gender should be sex Corrected as proposed Paragraph 4 line 2 & 3: was this explanation supported by your data: had the oldest old less chronic diseases? 5.4. Paragraph 4 line 9: utilization in these two settings : it was unclear which two setting were meant. I suppose the regulations are only different for living in the community versus nursing home? 5.5. Paragraph 4 line 11-13: On the other hand, these signs of a weak age-related increase in utilization may also be due to the crosssectional approach of our study, since high utilizers may die at an earlier age, leaving the cohort nearly unchanged with regard to utilization of services I don t understand what is meant with leaving the cohort nearly unchanged. It is a cross-sectional study so to population cannot change (in case of a cross-sectional approach the term cohort is not appropriate) No, the oldest old hat slightly more chronic conditions than the younger old. The corresponding data are presented at the bottom of this cover letter. One may see an indirect support in the fact that the difference in the number of chronic conditions between the youngest and the oldest age-group was only 1, although the difference in average age between the youngest and the oldest age group was 17 years.these data were presented as a graph in: van den Bussche H et al.: Which chronic diseases and disease patterns are specific for multimorbidity in the elderly? Results of a claims data based cross-sectional study in Germany. BMC Public Health 2011, 11:101. The point was included in the discussion section. The corresponding half sentence was omitted as unimportant As for the terms cohort and cross-sectional approach see heading 12. leaving the population nearly unchanged was an odd wording, and irrelevant by the way, and therefore omitted. 8

10 5.6. Paragraph 5 line 5 & 6: combining these figures seemed to me a Reworded in the discussion section 40 rather odd expression, certainly with the reference to previous work Paragraph 6 line 2: when controlling for other conditions. This Corrected as proposed way it seems that separate diseases (conditions) were controlled for in 41 the regression analyses. I prefer when controlling for other covariables such as nurse dependency Paragraph 8 line 1: comparatively high rate compared to what, to comparatively was omitted 42 whom? 5.9. Paragraph 9 line 12: the mean difference to the USA. Shouldn t it USA is systematically used. be the main difference? Also USA is inconsistently use e.g. in paragraph Mean was replaced by main line 1 the United States are mentioned. And what does this Last sentence was omitted sentence add to this paragraph? I believe it belongs to paragraph Paragraph 10 line 1 & 2: Interestingly, neither the profession nor This comment stems from daily observation of the socio-political scenery in the general public in Germany are aware of this distinctive situation. If Germany but not the result of a scientific study. Therefore, it was omitted. 44 they were, it would be generally considered to be self-evident and beneficial. Can you make these statements? Paragraph 11 line 8: spelling mistake seem play an important Correction: seem to play 6. Conclusions comparatively was omitted Line 6: comparatively compared to what? 7. Figure 1: Title now read mean number of contacts with physicians per year and mean It is not clear from the figure nor the title that the number of contacts is number of different physicians contacted per year 47 the mean number of contacts per year. (This applies also for the (mean?) number of physicians in figure 2) Table 1: gender should be sex Corrected as proposed Number of chronic conditions according to age group Multimorbid sample Non-multimorbid sample Age-group All Men Women All Men Women (2.3) 5.3 (2.3) 5.4 (2.3) 0.7 (0.8) 0.6 (0.8) 0.8 (0.8) (2.4) 5.6 (2.4) 5.7 (2.5) 0.8 (0.9) 0.8 (0.9) 0.8 (0.9) (2.4) 5.9 (2.6) 6.0 (2.7) 0.9 (0.9) 0.9 (0.9) 0.9 (0.9) (2.5) 6.1 (2.7) 6.2 (2.8) 1.0 (1.0) 0.9 (0.9) 1.0 (0.9) All (2.4) 5.7 (2.5) 5.9 (2.6) 0.8 (0.9) 0.7 (0.8) 0.8 (0.9) Source: van den Bussche H et al. : Which chronic diseases and disease patterns are specific for multimorbidity in the elderly? Results of a claims data based cross-sectional study in Germany. BMC Public Health 2011, 11:101. 9