Objectives: To identify a new scoring algorithm for the M-CHAT maintaining high sensitivity and improving PPV by reducing false positives and reducing the number of Follow-up Interviews needed.
Methods: The sample included 15,650 toddlers screened during checkups (mean age=20.6 months, SD=3.1, range=14-30 months; 7804 males). Potential scoring algorithms were developed based on item analysis, M-CHAT-to-Interview change scores, and discriminant function analysis, and examined for psychometric properties. The new scoring algorithm, called M-CHAT Best7, consists of interest in peers, pretend play, protodeclarative point, shows objects, responds to name, follows point, and wondered if deaf.
Results: Using the original M-CHAT scoring (screen positive=2/6 critical or 3/23 total items), 8.7% (1359/15,650) of the sample screened positive. Of those reached for interview, 13.1% (141/1074) screened positive and were offered an evaluation. Of those children evaluated, 51.8% (58/112) were diagnosed with an ASD. An additional 29 cases were evaluated based on physician or parent concerns, leading to a total of 64 ASD and 77 nonASD cases. Thus, PPV based on M-CHAT+Interview is .52, and PPV for M-CHAT alone is .04. The upper bound of sensitivity, considering ASD cases flagged by physician or parent, is .91, and specificity is .99.
Using the M-CHAT Best7, 1.9% (304/15,650) screened positive, and 42.5% (90/212) screened positive on the Best7 Follow-up Interview. Had we used Best7 initially, 55 ASD cases would have been identified, leading to an upper bound of sensitivity of .86, specificity of .99, PPV of .61 for M-CHAT+Interview, and PPV of .18 for M-CHAT only. For children whose Best7 score was ³4, the likelihood of reversing a screen positive score on Interview was small (22.4%), in contrast to the likelihood of reversing scores of 2 (81.4%) or 3 (43.1%).
Conclusions: Although the M-CHAT Follow-up Interview is still needed, the new Best7 scoring reduces the false positive rate fourfold, with only minimal decrease in sensitivity. On initial screening, Best7 scores 0-2 indicate low risk and no follow-up is necessary unless surveillance or other procedures suggest risk for ASD. The M-CHAT Follow-up Interview should be administered for Best7 scores of 2-3, and cases who screen positive on the Interview should be referred for diagnostic evaluations and early intervention. For cases whose Best7 score is >3, the Interview can be bypassed and referrals for evaluation and intervention made immediately. This recommended protocol retains high sensitivity and significantly reduces the false positive rate, making the best use of professional resources.