Reference standard diagnoses were established by a resident physician (emergency medicine) rating the 883 charts according to specific criteria, reported previously12 and summarized in the report’s Appendix.* The rater classified patients by their primary/predominant problem for the entire year as organic disease, somatization, or minor acute illness. The designation of the primary problem was based on the largest number of visits for a problem. For example, a patient with a documented urinary infection at the first visit with 1 follow-up visit, documented pneumonia at the third visit with 2 follow-up visits, 1 visit for chronic low back pain with a negative computed tomography scan, and 2 visits for minor ligamentous strain, with no objective manifestations and no investigation, would be rated as organic for this year; similarly, a patient would be considered to have minor acute illness with 6 visits for minor complaints with no work-up and no objective manifestations of disease on examination, as well as 1 visit for documented urinary infection and 3 for diabetes mellitus. The same rule was used for follow-up ratings 1 and 2 years later, and the rater for follow-up ratings was unaware of the baseline ratings.
Organic disease was diagnosed by standard medical criteria and based on clear physical signs of disease (eg, laceration, enlarged liver) or, almost always, definitive laboratory investigation; the rater relied on expert judgment and referred to text material as needed.13 Somatization was rated when, following objectively based diagnostic evaluation (definitive testing), patients were free of organic disease that contributed significantly to at least 1 physical symptom of at least 6 months’ duration. Minor acute illness was rated when all physical symptoms were of less than 6 months’ duration, as judged by the rater from explicit mention in the chart or from observation that symptoms cleared and did not recur, and there was no documentation of an organic disease explanation for the symptom or its degree of severity. Because minor acute problems typically were not severe or disabling (in contrast to somatization), definitive testing often had not been performed. From the 1995 baseline sample of 122 somatizers and 450 minor acute patients identified by our rating procedure, we re-rated a sample of all available somatizers (N=104; 85%) and a 15% random sample of minor acute illness patients (N=66) both 1 and 2 years later.
After 10 hours of initial training, including practice rating on nonstudy charts, the rater rated 20 charts of non-study high-utilizing patients. A priori, we set an agreement rate with the trainer (one of the authors [R.C.S.]) for primary diagnosis of 90% (18 of 20 charts) before the rater began rating study patients. During the study, the trainer rated sets of 20 study charts already evaluated by the rater, once each during 1995, 1996, and 1997. The rater had high levels of agreement with the trainer throughout, varying from 90% to 95%. This level of agreement is not surprising, because the trainer trained the rater, which was also reflected in the of 0.93.
Statistical analysis
We also reviewed the 1996 and 1997 use for the same 104 somatizers identified in our initial 1995 baseline chart review of high-utilizing patients. Again, the same 15% random sample (N=66) was selected from those classified as having minor acute illness in 1995. In the follow-up years, some patients had relocated or were no longer receiving their medical care at our HMO. However, nearly 85% of the patients in our selected sample were continuously enrolled in the HMO in the 2 subsequent years of our study. Similar to chart rating, the final sample consisted of 104 somatizers and 66 patients with minor acute illness. For these patients we ascertained their status (somatization, organic disease, or minor acute illness) and their use (<6 visits; 6 visits) in 1996 and 1997. The 2 groups of patients were compared using chi-square tests for categoric variables and by t tests for continuous variables. Confidence intervals for binomial proportions were calculated by the exact method.
Results
The characteristics of 170 patients (104 somatization; 66 minor acute) studied at all data collection points are shown in Table 1. The mean age (as of 1995) was 41.3 years among somatizers versus 39.7 years among patients with minor acute illness (P =.19) The 2 groups differed only by sex, with nearly 83% of the somatizers being women, compared with 65% among minor acute illness patients (P=.009).
Among 883 high-use patients at baseline, 311 had organic diseases (35%); 122 had somatization (14%); and 450 had minor acute illness (51%) as their primary problems. No patients with initial minor acute illness diagnoses were reclassified as somatization 1 and 2 years later, and all but 2 patients had minor acute illness in 1 or both follow-up years.