Methods

Data Extraction

The initial LIS data extraction generated a dataset containing information relating to all pathology tests conducted on specimens received by the pathology service departments in the period January 2008 and September 2011. Our analyses were conducted only on pathology tests that were ordered by the six study hospitals. The analysis was focused by further limiting the dataset to pathology tests conducted on specimens received during August and September for each year: 2008, 2009, 2010, and 2011. This reduced dataset contained information relating to 3,227,101 pathology tests. Within the dataset, 429,068 (13.3%) records were found to be duplicate entries (where the values in every field were identical). Once duplicate records were removed, the dataset contained information for 2,798,033 tests. A further 30,359 records were removed because they related to laboratory workflow rather than identifying an actual test order. This left 2,767,674 pathology test records associated with 130,060 patient records (who may have had multiple admissions in hospital). This dataset formed the basis for the subsequent analysis of test volume and turnaround times. Another adjustment was made to these data to account for a small proportion of tests whose turnaround time was recorded with a value less than zero minutes (for data entry time, 10,474 such records were found; for Total Laboratory TAT, 890 such records were found). These records were flagged and did not contribute to analyses of TATs, but were included in other analyses.

In order to assess the volume of test ordering per patient encounter (from patient admission to the hospital until their discharge) it was necessary to extract patient encounter data from the Patient Administration System (PAS) and Emergency Department Information System (EDIS) of the hospitals. These patient encounter data covered the period between 1 August and 30 September of 2008, 2009, 2010, and 2011. A number of steps were taken to ensure the integrity and consistency of these patient encounter datasets before they were linked to the test order dataset. The final linkage occurred between records for 147,280 patient admissions (extracted from the PAS), and records for 176,015 ED presentations (extracted from the EDIS), with the records for 2,767,674 pathology test orders (extracted from the LIS).

Data Linkage

All data integrity and validity checks, and linkage were performed in IBM SPSS Statistics 20.0.0. The datasets extracted from the PAS and EDIS were comma-separated values (CSV) format; the in-built SPSS data opening functions were used to import the data.

The patient admission dataset from the PAS and the ED presentation dataset from the EDIS were merged with the Test Order dataset from the LIS and the entire merged dataset was sorted by patient, patient admission dates and times, and specimen collection dates and times. Test orders where the specimen was collected after the patient admission and before the patient discharge, for matching patients, could be confidently attributed to those patient encounters. Data linkage between the three datasets allowed a single test order to be linked with either the PAS or EDIS dataset, or both datasets simultaneously. The SPSS “LAG” function was used to compare the patient, patient admission dates/times, and specimen collection dates/times of the sorted merged datasets and to associate, where valid and appropriate data were found, patient admission, discharge, and demographic information with the relevant test order data. In cases where specimen collection for a test order occurred either before patient admission, after patient discharge, or where no patient encounter data could be found, no linkage was performed. Therefore, these test orders were excluded from all analyses where linked data were necessary (e.g., comparisons of test rates per patient admission and DRG casemix). Once the linkable patient presentation and admission data from the EDIS and PAS datasets were merged, the merged dataset was cleaned to remove orphan patient admission information (presentations and admissions for which no associated pathology tests were found).

Data Analysis

Data analyses were conducted using IBM SPSS Statistics 20.0.0 and Microsoft Excel 2007. A number of different statistical tests were used for tests of significance. These depended on the nature of the data being analysed, and the research question being addressed. At various points of this report, analyses used independent-sample t-tests, chi-square (χ2) tests of independence, Mann-Whitney U tests, and Wilcoxon signed-rank tests. In all cases, the alpha-value for significance was set at p < .05.