Excluding deep vein thrombosis safely in primary care

In the validation population, only 2 patients in the “low risk” category (score 4–6) had DVT (TABLE 2). Both patients scored 6 points on the rule. That DVT was absent in all patients with a score of ≤5 strongly confirms the safety of the chosen threshold (ie, not referring patients with a score of ≤3).

Like in the original study, we presented the distribution of patients over 4 risk categories, with accompanying diagnostic accuracy measures (TABLE 2). With increasing scores, the risk of DVT increased. When the threshold of not referring a patient would be raised, sensitivity decreases (more false negatives) and specificity increases (more true negatives). In other words, by increasing the threshold, fewer patients will be referred (saving referrals) at the cost of a higher percentage of missed DVT cases.

Referral still a judgment call. We recommend using the threshold of ≤3 when the rule is applied. However, use of the rule is discretionary, not mandatory. The clinical view of the physician remains important; one may still decide to refer a patient with a “very low risk” if the suspicion of DVT remains. On the other hand, circumstances may prompt a wait-and-see decision even for a patient with a score of 5.

Two caveats. First, rigorously derived prediction rules may lose their accuracy when applied to other settings, because predictors may be idiosyncratic to the population in which the rule was developed.^5,7,8 Further research could focus on whether our rule yields similar diagnostic accuracy in other settings, including secondary outpatient care.

Second, we imputed missing values in this study. Although it is acknowledged that imputing missing values is better than simply deleting all patients with one or more missing values, we repeated the entire analysis on the complete cases. The complete case analysis did not yield different results than the analysis on the imputed data.

In conclusion, this validation study demonstrated that, also in a new patient sample, the primary care physician can safely refrain from referring a considerable number of patients suspected of DVT by using a simple diagnostic rule. The use of this rule reduces the number of unnecessary patient referrals to secondary care and consequently patient burden.

CORRESPONDENCE
Diane Toll, MSc, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands, P.O. Box 85500, 3508 GA Utrecht, The Netherlands. E-mail: D.B.Toll@umcutrecht.nl