| Table 3: Potential limitations of studies using secondary sources of data | |
| Feature | Potential limitations |
| Accuracy and precision of data |
Basic information on patient characteristics, diagnoses and procedures, particularly subsidiary diagnoses and secondary procedures, may be miscoded
Data generally available only for people who use services or access health care system in the period of interest Patients may be miscounted if they have multiple sites of residence or have undergone procedures out of province |
| Sample size | Large samples in administrative data raise risk of statistical significance without clinical or policy significance* |
| Characterization of patients | Patients are incompletely characterized; key clinical data on processes and outcomes may be missing |
| Characterization of providers | Identities and characteristics of providers are inconsistently recorded |
| Characterization of interventions | Services or procedures may be miscoded, or they may be coded accurately but according to broad definitions that are not clinically precise |
| Ascertainment of outcomes | Outcomes of greatest interest may not be captured (e.g., symptom relief, quality of life, out-of-hospital events and level of satisfaction) |
| Completeness of follow-up | Administrative data and most registries track only discrete episodes of service or care. Longitudinal follow-up requires data linkage, which may not always be possible |
| *Clinical significance refers to effects that might be judged important by clinicians or patients; policy significance refers to findings that might be judged worthy of action by policy-makers or administrators. | |