| Dutch guideline | AHCPR | VA | |
|---|---|---|---|
| History assessment | |||
| Presenting problem | R | R | R |
| Past medical history | R | R | R |
| Medication use | R | R | R |
| Drug/alcohol use | R | R | R |
| Family medical history | R | R | R |
| Personality change | R | — | R |
| Report of daily activities | R | — | R |
| Suicidality | — | — | (R) |
| Clinical examination | |||
| Physical | R | R | R |
| Neurologic | R | R | R |
| Sensory | R | R | R |
| Cognitive status | R | R | R |
| Speech/language | R | — | — |
| Depression | R | R | R |
| Delirium | R | R | R |
| Neuropsychological testing | T | T | T |
| Hachinski score | — | — | — |
| Psychiatric assessment | — | — | R |
| Psychotic symptoms | — | — | R |
| Standardized functional assessment | — | R | Opt |
| Additional tests | |||
| Hematology (Hgb, HCT, MCV, ESR) | R | R | R |
| Biochemistry (glucose, creatine, thyroid function) | R | R | R |
| Kalium | T | R | R |
| Liver function (ALAT) | T | R | R |
| Liver function specific (GGT) | T | R | R |
| Vitamin B12 | T | — | — |
| Folate acid | T | R | — |
| Brain scanning | Opt | Opt | T |
| Human immunodeficiency virus | — | T | T |
| Toxicology screen | — | T | T |
| Urinalysis | — | R | — |
| Drug concentrations | — | — | — |
| Albumin | — | — | — |
| Cerebrospinal fluid analysis | — | T | T |
| Electrocardiogram | — | R | — |
| Electroencephalogram | — | T | T |
| Chest x-ray | — | R | — |
| Specialist consultation and/or referral | Doubt about dementia or its cause | with mixed results | — |
| AHCPR, Agency for Health Care Policy and Research; ALAT, alanine aminotransferase; ESR, erythrocyte sedimentation rate; GGT, gamma-glutamyltransferase; HCT, hematocrit; Hgb, hemoglobin; MCV, mean corpuscular volume; Opt, optional; R, recommendation; (R), implied recommendation; T, targeted; VA, US Department of Veterans Affairs. | |||
Measurements
For every patient suspected of having dementia based on the DCGP guideline, the family physicians recorded their assessment findings, whether a close relative was available, the presence of dementia (yes, no, unsure), the final diagnosis, the number of contacts needed, and their actions taken to reach the diagnosis. Assessment of cognitive functions, behavioral changes, and somatic comorbidity were scored trichotomously: normal, unsure, or impaired. The level of ADL dependency of a patient was scored on a 4-point Likert scale ranging from independent to severely dependent. For cognitive disorders, behavioral changes, and comorbidity, sum scores were computed to reduce the number of variables. A sum score of 11 variables was made for cognitive disorders (Cronbach’s a = 0.75): long- and short-term memory; orientation to time, person and place; praxis; gnosis; language ability; abstraction; judgment; and personality changes. A sum score of 6 variables was made for behavioral changes (Cronbach’s a = 0.65): aggression, apathy, restlessness, denial, depression, and incontinence. Finally, a sum score of 5 variables was made for comorbidity (Cronbach’s a = 0.78): internal (medical) dysfunction, neurologic dysfunction, sensory impairment, adverse effects, and drug intoxication.
A sum score was made of the number of recommendations made from a list of 31 possible recommendations (Cronbach’s a = 0.76). Two indicators for continuity of care, namely, the length and familiarity of the family physician–patient relationship, were recorded by the family physician on a 4-point Likert scale.
DSM-III-R diagnoses
In addition to the diagnosis by the family physician, we also determined the diagnosis based on the DSM-III-R criteria. The findings recorded by the family physician were applied independently by 2 researchers blinded to the DSM-III-R criteria. Differences were discussed and consensus was reached in all cases.
Reference standard: memory clinic
An experienced multidisciplinary team that included a geriatrician, a neurologist, and a psychologist assessed the presence of dementia in all suspected referred patients. The memory clinic’s team was blinded to the family physicians’ and DSM-III-R diagnoses. To this aim, the CAMDEX (Cambridge Mental Disorders of the Elderly Examination)23 and the criteria of the DSM-IV were used.24 Studies on diagnostic accuracy of memory clinic teams compared with postmortem diagnostics show high levels of accuracy (80%–90% of diagnostic agreement).25,26 To our knowledge, the inter- or intraobserver reliability of memory clinic diagnoses has not been studied. The assessment, interpretation, and communication of the results took approximately 4.5 hours spread over 3 visits.
Statistical analyses
The accuracy of the family physicians’ diagnoses and DSM-III-R diagnoses in comparison with the memory clinic diagnosis (reference standard) was estimated using the sensitivity, specificity, positive and negative predictive values, and likelihood ratios. Univariate logistic regression analyses were used to quantify the association of clinical and sociodemographic characteristics, continuity of care, and performance indicators with the presence and absence of dementia. All determinants with P 27-30 The overall model was then reduced by excluding variables with P > .05 to obtain a simpler diagnostic model. The reliability of the overall and reduced diagnostic model was assessed by using the Hosmer and Lemeshow test.27
The ability of the overall and reduced model to discriminate between patients with and without dementia was quantified using the area under the receiver-operating curve (ROC area).31 The area under the ROC curve is a measure of the ability of a test to discriminate between patients with and without a disease, and can range from 0.5 (no discrimination, like flipping a coin) to 1.0 (perfect discrimination). A value between 0.7 and 0.8 is considered to represent reasonable discrimination, and a value of more than 0.8 is good discrimination.32 Differences in diagnostic discriminative value between different models and variables were estimated by comparing ROC areas, taking into account the correlation between models as they were based on the same cases.33 To perform these analyses, the family physicians’ diagnoses had to be dichotomized intodementia present and absent. We chose to classify the 8 cases with family physicians’ diagnoses “unsure” as dementia absent, as the DCGP guideline recommends a reluctant policy in such cases. We performed a sensitivity analysis to check whether the classification of these 8 cases as “dementia present” would have resulted in different findings.