Dementia: Predictors of diagnostic accuracy and the contribution of diagnostic recommendations
- OBJECTIVES: To explore and quantify the relative contribution of guideline recommendations and other determinants in the family physician’s diagnostic work-up of patients suspected of dementia.
- STUDY DESIGN: We prospectively studied 64 family physicians in an Eastern district in the Netherlands who diagnosed dementia according to the national Dutch guidelines in primary care. Their diagnoses were compared with the reference standard embodied by the memory clinic team of the University Medical Center Nijmegen.
- POPULATION: The physicians evaluated 107 patients older than 55 years suspected of having dementia.
- OUTCOMES MEASURED: Predictive value of various clinical and demographic parameters were measured in both univariate and multivariate logistic regression analyses.
- RESULTS: Activities of daily living (ADL) dependency (odds ratio [OR] = 5.3, P = .03), years since symptoms first started (OR = 1.84, P = .03), and the presence of somatic comorbidity (OR = 0.48, P = .02) independently contributed to the prediction of the presence or absence of dementia. The area under the receiver-operating characteristic (ROC) curve for these 3 variables together was 0.79. The ROC area of the family physicians’ diagnosis to determine the final diagnosis was 0.74. The number of recommendations applied did not additionally contribute to the assessment of the final diagnosis.
- CONCLUSIONS: The diagnostic accuracy of the family physician was reasonable. For family physicians, ADL dependency is a better predictor of dementia than cognitive impairment. Family physicians should be aware of diagnostic difficulties in patients with somatic comorbidity. We were unable to confirm the diagnostic value of many of the recommendations of dementia guidelines.
Results
Patients and family physicians
Over 16 months, 64 family physicians enrolled 107 patients suspected of having dementia, a mean of 1.7 patients per family physician. The participating family physicians were aged an average of 47 (SD = 7) years old and handling a practice population of 2113 (SD = 600) patients. Their characteristics were comparable to other Dutch family physicians except that they included fewer solo practitioners (32% versus 49% nationwide) and slightly more female family physicians (21% females versus 17% nationwide).34 Both the family physicians and the memory clinic completed the diagnostic evaluation for 93 patients: 14 patients dropped out because of refusal (n = 9), medical complications (n = 3), or death (n = 2). The clinical and demographic characteristics of these 14 patients were comparable to those of the 93 completed patients. Of the 93 patients, 93% lived independently and 62% were married. Other demographic characteristics are shown in Table 4. For 22 patients no informant was available (23.6%). The available informants were partners (77%), children or stepchildren (19%), or friends, neighbors, and others (4%). Of the informants, 67% were female and 66% shared a household with the patient.
A mean of 26 of the 31 recommendations (84%) was applied for each patient (SD = 3.3; range, 15–30). The family physicians needed on average 3.6 (SD = 3.3) contacts to assess a patient, and 40% received a home visit. Most patients were well known to the family physician; only 18% were not at all or only somewhat familiar. The MMSE was used as a diagnostic tool for only 19%. The mean time between the last assessment contact of the patient with the family physician and the first visit at the memory clinic was 61 days (SD = 39). The mean duration of the symptoms before the assessment was 22 months (SD = 13).
Diagnostic accuracy
Table 3 shows the accuracy of the family physicians’ and the DSM-III-R diagnoses compared with the memory clinic diagnoses. The prior probability of dementia was 63.4% (59/93). A positive diagnosis by the family physician increased the probability (positive predictive value) to 80.3%, and a negative diagnosis decreased this probability to 31.2% (10/32). The positive and negative predictive values of the DSM-III-R criteria were much lower (Table 2).
The 9 patients classified unsure by the family physicians were diagnosed by the memory clinic team as having amnestic syndrome (n = 3), dementia (n = 2), delirium (n = 1), age-dependent cognitive decline (n = 1), depression (n = 1), and unavailable (n = 1). Of the 12 patients with a false-positive diagnosis, 6 showed cognitive impairment, but did not fulfill all diagnostic criteria of dementia, and 1 patient received a diagnosis of depression. Of the 8 patients with false-negative findings, 6 had Alzheimer’s disease, 1 had dementia with unknown cause, and 1 had a normal pressure hydrocephalus. The family physicians expressed diagnostic confidence in 59% of all cases and in 47% of the patients diagnosed with Alzheimer’s disease.
Classification of the 8 patients labeled “unsure” as “dementia present” or “dementia not present” led to only small differences in the positive and negative predictive values (Table 2).
TABLE 2
Dementia diagnoses of family physicians and DSM-III-R compared with the memory clinic team (reference test)
| Reference test | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Dementia (n = 59) | No dementia (n = 34) | Total | PPV | NPV | SE | SP | LR+ | LR– | |
| Family physician diagnosis | |||||||||
| Dementia | 49 | 12 | 61 | 0.80* | 0.69* | 0.85* | 0.65* | 2.43* | 0.23 |
| Unsure | 2 | 6 | 8 | ||||||
| No dementia | 8 | 16 | 24 | 0.74† | 0.66† | 0.86† | 0.47† | 1.62† | 0.30† |
| DSM-III-R criteria‡ | |||||||||
| Dementia | 13 | 8 | 21 | 0.62 | 0.36 | 0.22 | 0.76 | 0.92 | 1.03 |
| No dementia | 46 | 26 | 72 | ||||||
| *Dichotomizing the family physicians’ diagnoses by grouping the unsure to the category “no dementia.” | |||||||||
| †Dichotomizing the family physicians’ diagnoses by grouping the unsure to the category “dementia.” | |||||||||
| ‡The registered symptoms were integrated by the researchers according to the DSM-III-R criteria. | |||||||||
| LR+, positive likelihood ratio; LR–, negative likelihood ratio; NPV, negative predictive value; PPV, positive predictive value; SE, sensitivity; SP, specificity. | |||||||||
Univariate associations of each documented variable by the family physicians
| Dementia | ||||
|---|---|---|---|---|
| Absent | Present | OR (95%CI) | P | |
| Clinical findings by family physician | ||||
| Cognitive symptoms | ||||
| (0–12), mean (SD) | 7.6 (5.2) | 9.8 (4.6) | 1.13 (1.03–1.24) | .01* |
| ADL dependency, % | 25 | 75 | 3.53 (1.46–8.56) | .01* |
| Somatic comorbidity, % † | 88 | 67 | 0.27(0.08–0.89) | .03* |
| Blood abnormality, % ‡ | 21 | 42 | 2.84 (1.07–7.55) | .04* |
| Behavioral changes, % | 41 | 68 | 0.74 (0.54–1.02) | .07* |
| Duration of symptoms, years (SD) | 1.5 (0.8) | 1.9 (1.0) | 1.77 (1.07–2.94) | .03* |
| Family physicians’ performance | ||||
| Number of consultations, mean (SD) | 3.9 (3.4) | 3.5 (3.6) | 0.97 (0.86–1.09) | .60 |
| Home visit, % | 35 | 43 | 1.39 (0.58–3.33) | .46 |
| Recommendations applied, mean (SD) | 24.5 (3.4) | 25.0 (3.7) | 1.04 (0.92–1.17) | .57 |
| MMSE used, % | 21 | 15 | 0.69 (0.23–2.07) | .51 |
| Informant contacted, % | 65 | 83 | 2.67 (1.01–7.11) | .05* |
| Familiar with patient, % | 79 | 84 | 1.41 (0.47–4.21) | .54 |
| Family physician–patient relation, >5 years, % | 65 | 75 | 0.60 (0.24–1.50) | .27 |
| Patient characteristics | ||||
| Mean age, years (SD) | 73 (8.7) | 74.3 (6.3) | 1.03 (0.97–1.09) | .41 |
| Male sex, % | 44 | 44 | 1.00 (0.42–2.34) | .99 |
| n = 93. Values are in means (SD) or percentages. | ||||
| *P | ||||
| † Sensory impairment, internal dysfunction; neurologic dysfunction; intoxication; adverse drug effect, sum score of dichotomous items. | ||||
| ‡ Hematology (hemoglobin; hematocrit; mean cell count; erythrocyte sedimentation rate); biochemistry (glucose; creatine; thyroid function), sumscore of dichotomous items. | ||||
| ADL, activities of daily living, 1 question scored on a 4-point scale; CI, confidence interval; MMSE, Mini-Mental State Examination; OR, odds ratio. | ||||
