Conformance Verification of Healthcare Application Code Against Clinical Guidelines: A Case Study of Blood Pressure Classification

Healthcare applications increasingly implement clinical guidelines as software-based decision rules. However, validation of measurement accuracy does not establish whether the implemented classification logic conforms to the corresponding clinical guideline. This study evaluated whether exhaustive conformance verification can identify discrepancies between healthcare application code and a named clinical specification. The ۲۰۲۳ European Society of Hypertension (ESH) guideline was formalized as a deterministic reference oracle for blood pressure classification. An exhaustive integer input space comprising systolic blood pressure values from ۶۰ to ۲۶۰ mmHg and diastolic blood pressure values from ۴۰ to ۱۶۰ mmHg was generated, yielding ۲۴,۳۲۱ ordered input pairs. Fifty-seven open-source blood pressure classification applications were evaluated against the reference oracle. Binary hypertensive status and, where applicable, multi-category classification were assessed independently. Disagreements were traced to the corresponding decision branches and characterized according to their structural pattern and clinical direction. All ۵۷ applications reproduced the binary ESH hypertensive/non-hypertensive decision across the complete input space, resulting in ۱۰۰.۰۰۰% sensitivity, specificity, and accuracy. Category-level analysis was possible for ۴۱ applications. Fifteen applications achieved ≥۹۹.۹% conformance, ۱۰ achieved ۸۰–۹۹.۹% conformance, and ۱۶ achieved <۸۰% conformance. The lowest category-level conformance was ۴.۳۵۴%. The identified discrepancies primarily involved under-classification and were associated with cascade-ordering errors, coverage gaps, and unreachable crisis branches. One implementation had ۲۳,۱۹۰ of ۲۴,۳۲۱ input combinations classified outside the applicable ESH categories. Binary agreement with a clinical guideline did not establish complete conformance at the category level. Exhaustive verification provided a deterministic and reproducible method for evaluating bounded clinical classification logic and for tracing discrepancies to specific implementation structures. This approach may complement measurement validation, usability assessment, security testing, and other forms of healthcare software assurance.

کلیدواژه ها:

Conformance verification ، Clinical guidelines ، Blood pressure classification ، Healthcare application ، ESH ۲۰۲۳