Abstract:
Background: Calculation of reference intervals (RI) for the population served is a standard
requirement to assure accurate interpretation of laboratory test results. Conventional method or
direct method of calculation of RI for laboratory parameters for elderly population is practically
not feasible. Literature recommends use of indirect method as an alternative. RI are not defined
for complete blood count (CBC) parameters of elderly population in Sri Lanka.
Objectives: To assess the plausibility of using patient data to calculate RI of CBC parameters of
elderly patients.
Methods: A retrospective, cross-sectional study was carried out using data of all the patients
>65 years stored during two consecutive years in a hospital laboratory. Reports with flagging
for any parameter, reports of inpatients and repeated samples and outliers in each parameter
identified using Tukey’s method were excluded. Distribution plots, Q-Q plots and box plots
were drawn for all refined data and visually inspected for their normality of distribution. Since
data were normally distributed, RI were defined as mean±2SD. Calculated RI were compared
with RI provided by manufacturer of the automated CBC analyser and with RI defined in five
other countries.
Results: There was a total of 3094 CBC reports of patients >65 years of age. Of those, only
1604 (51.8%) fulfilled the inclusion criteria. Distribution plots of data of all the parameters in
both male and female groups showed normal distribution. Calculated RI of all CBC parameters
were similar to those of other countries. RI obtained for red cell indices, platelet count and
white blood cells were similar to the RI provided by the manufacturer for adults. RI obtained
for red blood cell count, haemoglobin and packed cell volume were clinically significantly
lower compared to the RI provided by the manufacturer for adults.
Conclusions: When appropriate selection criteria and appropriate statistical tools are applied,
patient data can be used to calculate RI for CBC parameters in elderly people. Differences
observed with manufacturer defined RI confirm the need for defining laboratory’s own RI for
the populations served.
