Validation of LEST (0-3): The total sample for validation was calculated as follows. N (the number of positives for language delay) required was (1.96)2PQ/d2; where P is the sensitivity expected Q is (1-P) and d is the precision desired. Taking the expected sensitivity of the new tool as 95% and precision as 10 and with 95% confidence interval, the sample size calculated was 18 positive cases (of language delay). Assuming 3% prevalence for delay in language development in 0-3years of age, the total sample to be studied was calculated as 600. Considering possibility of non-co-operation or incomplete tests (either the new tool or the reference standard) as 5%, the final sample was calculated as 632 (rounded to 650). Thus, a total of 653 children between 0-3 years belonging to 11 Anganwadi areas from an urban ward, 20 from a rural Panchayat and 3 from tribal area, participated in the study. They were accompanied by a parent/primary care giver. Final valid sample (where the new tool and the reference standard were available) obtained for analysis was 643. Anganwadi worker recruited the mothers and children between 0-3 years of age in their area to the respective Anganwadi, on a specified date and time. The screening test and the ‘Reference Standard’ were administered by independent observers blinded to the results of the other tests. Administration procedures for the tests were standardised. Same raters assessed children in all the Anganwadis to maintain the uniformity in the assessment process after standardisation. Two raters administered LEST and another two assessors evaluated the children using REELS. They were trained in Child Development Centre (CDC) Thiruvananthapuram for applying the tools. LEST (0-3), was applied by trained persons having similar educational qualification as that of an ICDS Supervisor and REELS by Speech and language Therapists, after acquiring written consent from the parent and verbal assent (here only means ‘seeking co-operation’) from the child. All children who satisfied the selection criteria (with proper consent from parents or primary care givers) in the respective Anganwadi areas were included in the study. Exclusion criteria include those who were ill and uncooperative for testing.

A total of 643 children were included for the validation, where both results (LEST as the tool and REELS as the reference standard) were available, 194 children (45.4% boys) were below 12 months, 197 were between 13-24 month old (54.8% boys), and 152 were between 25-36 month old (57.1% boys). Web Table I shows the 33 items with the age range at which the items are to be achieved normally.

The test re-test reliability of the tool was done in a valid sample of 50 and it was found to be acceptable (intra-class correlation was 0.69, 95% CI: 0.46-0.83). The inter-rater reliability for the study was done in a valid sample of 50 and it was also found to be acceptable (intra-class correlation was 0.94 having a 95% CI: 0.9 – 0.97)

Table I shows the cross tabulation of LEST with one item delay taken as test positive (LEST delay) against REELS.

Table II shows the cross tabulation of LEST with two items delay taken as test positive (LEST delay) against REELS.

Table III shows the comparison of tool characteristics in 2 situations of new test (tool) criteria with one item delay as ‘LEST delay’ and two items delay considered as ‘LEST delay’.

Web Table II shows the tool characteristics at different age ranges.

TABLE I	LEST (0-3 Years) Against REELS (One Item Delay As ‘LEST Positive’)

TABLE  II	LEST (0-3 Years) Against REELS (Two Item Delay As ‘LEST Positive’)

TABLE III Test Characteristics With Two Different Criteria 

An Anganwadi based survey of developmental delay/disability in one ICDS block had observed that the speech and language delay was the commonest among developmental problems [9]. However, language develop-ment is not represented adequately in most developmental assessment tools. It was the felt-need to have a tool for assessing language delay, which can be used by health workers. A screening tool for language delay should be simple, less time consuming and easily understood by the community health worker and the parents. LEST (0-3) was designed to meet these requirements. It was important to validate LEST against the ‘best available’ assessment tool. Here we used the original version of REELS. A conscious decision was taken by us to include only the REELS criteria "delay" to be taken as the reference standard positive to have a better specificity of the reference standard test. Problems with REELS in the community setting are that it is a time consuming test and difficult to administer in a community setting; and it can be administered only by speech and language pathologists. LEST, on the other hand can be administered by any person with minimal training. LEST is easy to administer, items are simple to perform and is in Chart form, which is easier than the former one. REELS is in compound and complex language which is difficult to understand and apply in the community, but LEST is in a simple language.

Every possible effort was taken to avoid bias in this study. The Anganwadi workers recruited all children in the Anganwadi area and ‘all consented’ were participants to the study. Thus ‘selection bias’ was minimised. Tests were done by equally trained persons (with educational qualifications similar to ICDS supervisor) not familiar with the children in the area (where as Anganwadi workers and ICDS workers were familiar with the children). All were given clear instructions to administer the tests and record the result systematically. Uniform hands-on training were given to all investigators. The observer who administered and interpreted the REELS did not know the screening test results and vice versa. Thus measurement bias was minimised.

Shifting the test positivity (tool positivity) from one item delay to two items delay (LEST positive), resulted in a drop of the sensitivity from 95.8% to 66.7% though test specificity showed an increase (from 77.5% to 94.8%). For using LEST as a screening tool for delay in Language, we selected two item delay as test positive accepting a lesser (out of the two options) sensitivity of 66.7% because of the relatively higher Positive Predictive Value (PPV of 14.2% and 33.3%, respectively for one item delay and two item delay as the tool criteria positive) and lesser false positives in the screened sample. This option gives an excellent Negative Predictive Value (NPV) of 98.7% also, which is desirable for a screening tool. The choice between one-item or two-item delay in LEST depends on the need in the community, whether to have a highly sensitive test with a very low PPV or to have a lesser sensitivity with a relatively higher and acceptable PPV (but still low due to the low prevalence of the condition to be screened).Unlike sensitivity and specificity, the PPV and NPV are more influenced by the prevalence of the disease. Among these measures, both the Likelihood ratios are least dependent on prevalence. Here we are getting a LR positive of 12.9 (95% CI: 8.3 - 20.01) with two- item delay as tool positive criteria.

The limitation of our study is the use of an imperfect reference standard, which we were forced to accept because of the non-availability of a Gold standard for language delay for the age group to be screened. In the case of language delay, it is difficult to get a gold standard and the original REELS was used for validation. In epidemiological studies, in the absence of a gold standard the researcher may be forced to take an ‘imperfect’ gold standard. In such situations researchers may utilise proximate measures of the ‘gold standard’ as the criterion to assess validity. In these situations the kappa statistics is commonly used to assess agreement for estimating "validity". Here we calculated the Prevalence and Bias Adjusted Kappa (PABK) of LEST with REELS and was acceptable (0.88) with two-item delay as tool positive criteria.

The observation that the prevalence of Language delay (as per the screening tool) coming down from 25.2% to 7.5% (due to drop in false positives) when shifting from one item delay to two items delay taken as the criteria for ‘LEST delay’, has implications in planning large scale community level language development assessments and for interventional programs.

Likelihood ratio positive and likelihood ratio negative, which are relatively independent of the prevalence of the condition, are also found to be good for LEST, and hence considered acceptable. The test-retest reliability here was a little low, but acceptable. This is explained by a recall effect (parents reporting the items better in the second time or due to the child performing the items better in the second time when examined by the same observer who have already established a rapport during the first examination, even though there was a gap of two weeks between measurements). The inter-rater ICC was 0.94, which was fairly high.

LEST (0-3 years) is a valid Indian tool for identifying children of 0-3 years with language delay in the community with an acceptable Sensitivity, Specificity, Positive Predictive Value and Likelihood Ratios. This is a simple tool also, which can be finished in 10 minutes, by a health worker, and requires only a pen/pencil/scale along with the tool and minimal training to apply the tool. Based on the test result, language stimulation by the mother can be easily done at home and also help in referring needy ones to an appropriate referral centre for intervention.

1. Asher SR, SL Oden. Childrens’ failure to communicate. An assessment of comparison and egocentrism explanations. Developmental Psychol. 1976;12:132-9.

2. Jakobson R. Verbal communication. Scientific American. 1972; 227:73-80.

3. Shriberg LD, Tomblin JB, McSweeny JL. Prevalence of speech delay in 6-year-old children and comorbidity with language impairment. J Speech Hearing Res. 1999;42: 1461-81.

4. Nair MKC, George B, Philip E. Trivandrum Developmental Screening Chart. Indian Pediatr. 1991;28:869-72.

5. Frankenburg WK, Dobbs JB. The Denver Developmental Screening Test. J Pediat. 1967;71:181–91.

6. Bzoch KR, League R. Receptive-expressive Emergent Language Scale. 1st edition. Gainseville: The Tree of Life Press; 1971.

7. Coplan J. Early Language Milestone Scale. Austin, TX: Pro-Ed Inc; 1993.

8. Rossetti L. The Rossetti Infant-Toddler Language Scale: A Measure of Communication and Interaction. East Moline, IL: LinguiSystems;1990.