Comments

Health Affairs encourages readers to engage in discussion via comments on our Web site.

• To RESPOND to a particular article: Click on the link "Submit a response to this article" in the box at the top right-hand corner of the article.
• To READ responses to a particular article: Click on the link "View responses" in the box at the top right-hand corner of the article.

Comments to:

Susan Dentzer From The Editor Health Affairs, January/February 2009; 28(1): w87-w88. [Extract] [Full Text] [PDF] [Table Of Contents] [Reprints & Permissions]

Comments:Submit a response to this article

Comments published:

Ecological Cautions

Thomas Ricketts, G. Mark Holmes   ( 4 December 2008 )

Doctors And Quality Of Care

Arvind R. Cavale, MD, FACE, FACP   ( 16 December 2008 )

Geographic Correlations And Causality

Jonathan S. Skinner   ( 5 January 2009 )

Ecological Cautions 4 December 2008
Thomas Ricketts, Deputy Director UNC-CH Sheps Center for Health Services Research, G. Mark Holmes Send comment to journal: Re: Ecological Cautions ricketts{at}schsr.unc.edu Thomas Ricketts, et al.	The set of articles that debate the relationship between physician supply, costs, and quality of care that were published December 4 tend to ignore the need to look closely and critically at these relationships and explore alternative ways to understand what is happening. The analyses in these articles, in earlier articles by the same authors, and in work by Starfield, Shi and others attempt to draw conclusions about the ecological relationship of rates and ratios of physicians at various levels of geography to rates of mortality, indirect summaries of quality, and other indirect indicators of access. That is a task that is perhaps harder to accomplish than implied in the articles. An overriding problem with any analysis of this type is that the aggregation of data into rates is usually done using some arbitrary unit of geography that more closely aligns with political boundaries than cultural, economic, or sociodemographic geographic constructs. The Dartmouth Atlas group has gone to great lengths to attempt to create aggregates that are meaningful and appropriate to the assessment of the effectiveness of medical care and the relationship of the structure and financing of medical care to outcomes. Those areas, whether Hospital Referral Regions (HRRs) or primary care service areas, help with ecological analyses but are still rough compromises for the creation of denominators. Areas like states, counties, regions, and smaller all generate error in the analysis of health systems in the United States, given that we do not restrict use or residence to specific areas. The creation of the HRRs in the Dartmouth Atlas was a process of constant compromise, as the areas did not conform neatly to fixed boundaries, and eventually represented an estimate of what might be a useful denominator. The use of state-level indicators is suggestive at best and ought to be viewed with a good deal of skepticism, and the reliability of the conclusions, well short of the aggressive claims of causality that they have prompted in the interpretation of the cost-quality data. No degree of statistical manipulation of the associations will help without admitting that there is some substantial distorting effect caused by people and systems crossing the boundaries and very substantial variability within the boundaries. By using analytical methods that loosen the constraints of boundaries, we have tried to replicate and advance some of the analysis of physician supply to outcomes.[1] The results of that analysis suggest that we may need to pay more attention to the variability of our correlations and look to determine if there are more "real" geographies of influence. It appears that primary care physician supply is more or less consistently associated with lower mortality in some regions but higher mortality in others. These regions do not follow more convenient political state boundaries and tend to cluster areas in different ways and suggest trends across space that are not constrained by "markets," or "service areas," or state and county jurisdictions. The several authors of these papers need to continue to develop their analyses to account for these geographic issues. We certainly acknowledge that the preponderance of readily available data tends to be defined by political boundaries, often state-level, and the lack of data useful for answering interesting questions is a very real inhibitor in this line of inquiry. The development of more extensive quality surveillance systems will be useful in generating helpful data. Suggestive and stimulating as they may be, the existing ecological analyses are far from providing conclusive guidance for what we are to do in policy. NOTE Ricketts TC, Holmes GM. (2007) Mortality and physician supply, does region hold the key to the paradox. Health Services Research, Volume 42 No. 5. 2233-2250.
Doctors And Quality Of Care 16 December 2008
Arvind R. Cavale, MD, FACE, FACP, Private Practice Physician Feasterville, PA Send comment to journal: Re: Doctors And Quality Of Care mdllcoffice{at}gmail.com Arvind R. Cavale, MD, FACE, FACP	I find this discussion fascinating but not very conclusive. Given the wide disparity of views on this subject, it is very easy for policymakers to pick and choose which theory they would like to base their policies on. From a practicing physician's perpective, it would be safe to admit that high-quality care can come only from high-quality doctors, whether generalists or specialists. And just like any other service, one only gets what one pays for, in general. It is rather foolish to debate, as vigorously as the Harvard-Dartmouth group does, that one is better than the other. As far as your embargo policy goes, it is important to maintain it. However, barring scholars like Cooper from participating in such debates would tend to skew the discussion and make it rather imbalanced. Therefore, a more "practical" approach might be tried in the future.
Geographic Correlations And Causality 5 January 2009
Jonathan S. Skinner, Professor Dartmouth College Send comment to journal: Re: Geographic Correlations And Causality jon.skinner{at}dartmouth.edu Jonathan S. Skinner	I agree with Ricketts and Holmes when they note the importance of defining the right geographical regions in studying the association between physician supply and health outcomes. However, even perfect geographic areas won't solve the problem of untangling causality from simple correlations in cross-sectional aggregated analysis. So why do cross-sectional analysis at all? My view is that it provides a reasonable check on our theories and models. Suppose one’s view of the world is that more specialists (or health care intensity) lead to better health outcomes. But if we find in the real world a zero association (or positive in some areas while negative in others), then it suggests going back to the drawing board, and making an attempt to explain why we should observe the correlation we do. In other words, these associations can be used to test hypotheses but can never be used to make inferences about causal pathways. By the same token, I don’t find the association between the proportion of specialists and health care quality to say anything about the causality of whether specialists are “better” than primary care physicians, and thus I am sympathetic to Cavale’s concerns about the interpretation of the results.* If after this exchange we can agree that the empirical record indicates a positive association between generalists and quality, and an essentially zero association between specialists and process quality, then we can also agree that any theory or model of physician workforce should at least attempt to explain why this is so. * Some studies using good statistical methods suggest that specialists are quicker to adopt some important innovations; for example, see Fendrick, AM; Hirth, RA; Chernew, ME. Differences between generalist and specialist physicians regarding Helicobacter pylori and peptic ulcer disease. Am J Gastroenterol. 1996 Aug;91(8):1544–1548.