Table 1. Comparison of laboratory investigations using the conventional approach and the LAS (Clin Chem 45:8 1168-1175 (1999) Conventional LAS 32.7 t = 5.4, P <0.01 7.5 t = 3.4, P <0.02 17.8 $232 t = 3.3, P <0.05 3.2 66% 12 $194 1.7 12 1.7 8 No established guidelines: hepatitis and autoimmune testing Mean number of tests ordered 6.7 Total number of sample collections 38 3.3 30 Results returned negative Mean number of tests ordered Total number of sample collections 2 9 Results Mean number of tests ordered t-Test for two related samples Mean number of samples collected t-Test for two related samples Observations Mean costs, $ t-Test for two related samples Turnaround time (mean no. of days) Diagnostic accuracy Number of referrals Established guidelines: thyroid testing Mean number of tests ordered Total number of sample collections 4.8 14 5.8 1 100% 0 Enter Data Generate Rule Search Rule List NEW RULE Enter new Comment RULE EXISTS Search comment List Update rule list Retrieve Comment Update comment list Modify if required Search Rule List Variables Used TFT application • Qualitative – Gender – Clinical details – Drug therapy • Quantitative – – – – – Age Total T4 TSH Free T4 Free T3 Adult decision ranges TSH (mu/L) • Decision Range – – – – – – – – >=60 >=20; <60 >=10; <20 >5; <10 >4; <=5 >=0.4; <4 >=0.25; <0.4 <0.25 • Description – – – – – – – – Markedly high Very High High Mildly raised Near URL Normal Near LRL Below detection limit Clinical and Drug details TFT • CLINICAL DETAILS – A. Suspected hyperthyroidism – B. Suspected hypothyroidism – C.Post Radioiodine therapy – D. Post thyroid surgery – E. On thyroxine treatment – F. On T3 treatment – G. On Carbimazole – H. On PTU – Z. Non-specific • DRUG THERAPY – – – – – – – – A. Lithium B. Oestrogens C. Amiodarone D. Beta-blockers E. Glucocorticoids F. Phenytoin G. Carbamazepine H. Androgens – Z. No relevant drugs Variables Used Protein electrophoresis • Quantitative – patients age – serum total protein, albumin, -globulin, paraprotein concentrations • Semi-quantitative – Visual assessment of albumin, 1,,2,, 1,,,,2,, and concentrations • Qualitative – patients gender – Visual inspection of zones and interzones (11 variables), e.g. 1 zone, - interzone findings Primer knowledge bases TFT application Name N1 N6 N15 N16 Etc… TOTAL Cases 44 44 143 143 1142 Case Types Euthyroid Euthyroid (post RAI) Suppressed TSH (thyrotoxicosis) High TSH (Hypothyroid) Primer knowledge bases protein electrophoresis Name S1 S2 S3 S4 Etc… TOTAL Cases 6 18 32 34 206 Case Types Normal patterns Chronic inflammation patterns Beat-gamma fusion Altered alb/globulin fractions Performance evaluation - TFT Percentage AAI* per batch Percentage of automatically assigned interpretations (AAI) per TFT batch 100 90 80 70 60 Consecutive TFT batches TFT application statistics • • • • • • Number of variables specified: 8 Number of options/decision levels: 70 Number of theoretical data combinations: 20,412,000 Number of cases entered via primer knowledge bases: 1142 Number of rules: >2900 Percentage of automatically assigned interpretations [using the primer knowledge bases] – At the outset: 66% – After the first 500 patient cases: 70% – Maximum achieved: 95% • Number of interpretations rejected: <1% electrophoresis application statistics • • • • • • Number of variables specified: 23 (22 used in rule generation) Number of options/decision levels: 183 Number of theoretical data combinations: 8,800,000,000,000,000,000 Number of cases entered via primer knowledge bases: 206 Number of rules: 350 Percentage of automatically assigned interpretations [using the primer knowledge bases] – At the outset: 78% – After the first 500 patient cases: 72% – Maximum achieved: 70-80% • Number of interpretations rejected: 0 Clinical evaluation TFT application (27 TFT requests on 15 randomly-selected new referrals to a thyroid clinic were studied; 2 evaluators participated) Do you agree with the thyrometabolic classification in the report? Complete agreement Partial agreement Complete disagreement Evaluator 1 (physician) Evaluator 2 (biochemist) 96% 4% 0% 85% 11% 4% Was the interpretation clear in style 81% and presentation? 96% Were any references in the interpretation to the following items appropriate? Specific thyroid diagnoses [5] Interfering conditions/drugs [4] Further tests on the sample [1] Followup recommendations [16] Treatment recommendations [2] Actual no. of ‘No’ answers Actual no. of ‘No’ answers 0 4 0 0 0 1 0 1 1 0 Clinical evaluation Electrophoresis application (115 questionnaires circulated, 64 returned, 56% response) Question Answer - YES Answer - NO 1. Did you require an interpretation? 79% 21% 95% 5% 2. Was the interpretation helpful? 3. Did the interpretation confirm an existing or suspected diagnosis? 24% 4. Did the interpretation lead to a new diagnosis not previously suspected? 0% 5. Did you perform further investigations on the basis of the 14% interpretation? 76% 100% 86% Clinical evaluation Electrophoresis application (115 questionnaires circulated, 64 returned, 56% response) Question Answer - YES Answer - NO 1. Did you require an interpretation? 79% 21% 95% 5% 2. Was the interpretation helpful? 3. Did the interpretation confirm an existing or suspected diagnosis? 24% 4. Did the interpretation lead to a new diagnosis not previously suspected? 0% 5. Did you perform further investigations on the basis of the 14% interpretation? 76% 100% 86% Performance evaluation Electrophoresis application % AAI per batch of 100 cases 80 60 Without primer With primer 40 20 0 100 200 300 400 500 Total Number of reports entered Performance evaluation Electrophoresis application (to demonstrate knowledge acquisition) % AAI per batch of 100 cases 80 60 All reports labelled reports 40 20 0 100 200 300 400 500 Total Number of reports entered Cooperativity DATA KBS Interpretation Diagnostic outcome Cooperativity KBS-1 KBS-2 KBS-3 KBS-4 Diagnostic outcome (1) KBS-i Comment Diagnostic outcome (2) Comment Diagnostic outcome (3) Comment Diagnostic outcome (4) Comment Integrating Comment Cooperativity Organ-related profiles • Renal profile – Na, K, Urea, Creatinine • Bone profile – Ca, Alb, Corr Ca, Phosphate, AlkPhos • Liver profile – Bili, Alb, T Protein, AST, Alk Phos, GGT • Cardiac profile • Lipid profile Cooperativity Decision Levels 160 mmol/L Plasma Sodium Concentration 150 mmol/L 140 mmol/L 130 mmol/L 120 mmol/L 110 mmol/L Rule Elements Rules and rule elements Diagnostic outcomes Comments R Tr L Tl B Tb C Tc Integrating rule and its elements i Age Sex Clin Ti Cooperativity Primer knowledge bases Universe KBS (Primers are subsets) PRIMER PRIMER PRIMER PRIMER Renal KBS Bone KBS Liver KBS Cardiac/ Lipid KBS Selected Diagnostic Outcomes • Renal profile – – – – – – – prerenal impairment Renal impairment hyponatraemia hypokalaemia hypernatraemia hyperkalaemia etc • Bone profile – osteomalacia pattern – renal osteomalacia – primary hyperparathyroidism – etc Profile interpretations • Renal – – – – Na 133 [Low] K 5.8 [Raised] Urea 30.5 [raised] Creatinine 400 [Raised] • Interpretation – Renal impairment with hyperkalaemia and mild hyponatraemia • Diagnostic outcome – renal impairment • Bone – – – – Ca 1.94 Corr Ca 2.04 [Low] Alb 35 [Normal] Phosphate 1.88 [High] Alk Phos 450 [High] • Interpretation – Hypocalcaemia, raised phosphate and Alk Phos: osteomalacia pattern • Diagnostic outcome – osteomalacia pattern Cooperative KBS output • Renal KBS comment – Renal impairment with hyperkalaemia and mild hyponatraemia. [DO= renal impairment] • Bone KBS comment – hypocalcaemia, raised phosphate and Alk Phos. DO = osteomalacia pattern. • Integrating KBS comment – These results are consistent with renal failure and associated renal osteodystrophy Decision Support Systems Support for Health Care Professionals Alarm/Alert Systems Alert A significant but mild trend in a variable, or the achievement of a specified level by the variable. Univariate Alerts: based on a single variable Multivariate Alerts: based on multiple variables Alarm A marked trend in single variable, or the achievement of a critical specified level by the variable, and/or the presence of several (multiple) alerts or multivariate alerts. Univariate Alarms: based on a single variable Multivariate Alarms: based on multiples variables Alarm/Alert Systems Methods Alarms/alerts based on Critical Values Alerts/alarms based on Critical Trends Mathematical/Statistical methods Biodynamic Models KBS/Temporal reasoning systems Know ledge-based (intelligent) alerts/alarms Cooperative systems Alarm/Alert Systems Based on Significant Values Based on Significant Trends Univariate Alerts If Sodium_Concentration <125 mmol/l Then Set Alert [Sodium_Concentration] ON; End; If Potassium_Trend >0.1 mmol/l/hr Then Set Alert [K_Trend] ON End; Alarm/Alert Systems Based on Significant Values Multivariate Alerts If Sodium< 125 and Water_Balance > 1000ml and Urea NOT RAISED Then Set Alert [Dilution_Hyponatraemia] ON End; Based on Significant Trends If Potassium_Trend > 0.1 mmol/l/hr and K_Infusion > 5 mmol/hr and Potassium_Concentration> 4.5 mmol/l Then Set Alert [Hyperkalaemia_ trend] ON End; Know ledgebased alerts If and and and and and Then End; Sodium< 125 mmol/l Urine Osmolality > 350 mosm/kg cvp NORMAL Urea NOT RAISED Euthyroid/EuAdrenal CARBAMAZEPINE Therapy Set Alerts [SIADH]; [CBZ effect] ON Alarm/Alert Systems Calibration Alarms, when triggered, should lead to an immediate change in patient care. Alarms which do not lead to a change in patient care (false alarms) should be suppressed on subsequent occasions, or converted to an alert status. Suppression could also be achieved by knowledge-based revision of the Trigger Thresholds. Repeated alarms may be ignored. Alarms need to be responsive to the clinical context Alarms should draw attention to new or strange findings. Alarm/Alert Systems Based on Significant Values Based on Significant Trends Univariate Alarms If Sodium_Concentration <120 mmol/l Then Set Alarm [Hyponatraemia] ON; End; If Potassium_Trend >0.5 mmol/l/hr Then Set Alarm [K_Trend] ON End; Alarm/Alert Systems Based on Significant Values Based on Significant Trends Multivariate Alarms If Sodium<120 and Water_Balance > 1000ml and Urea NOT RAISED Then Set Alarm [Dilution_Hyponatraemia] ON End; If and and Then Potassium_Trend >0.5 mmol/l/hr K_Infusion > 5 mmol/hr Potassium_Concentration>5.0 mmol/l Set Alarm [Hyperkalaemia_ trend] ON End; Knowledge-based alarms If and and and Then End; Sodium<115 Alert [SIADH] ON Alert [Sodium_Trend] POSITIVE Alert [Hyponatraemia] ON Suppress Alarm [Hyponatraemia] Alarm/Alert Systems Variables specified in the Alarms/Alerts System. Test* Type** Categories*** Renal Sub-system: Sodium (mmol/l) Potassium (mmol/l) Urea (mmol/l) Creatinine (mol/l) Bicarbonate (mmol/l) Chloride (mmol/l) Glucose (mmol/l) QN QN QN QN QN QN QN 20 20 20 20 20 5 8 Acid-Base Sub-system: Arterial blood pH Arterial blood pCO2 Arterial blood pO2 QN QN QN 5 5 9 Haemodynamic Sub-system: Central venous pressure (mmHg) Cardiac failure index Oedema index Temperature index Pulse rate Blood pressure index Weight change Urine Output Sodium balance (mmol/day) Sodium output (mmol/day) Potassium balance (mmol/day) Potassium output (mmol/day) Nitrogen balance (g/day) Nitrogen output (g/day) Water balance (litres/day) QN QL QL QL QN QL QN QN QN QN QN QN QN QN QN 4 2 2 3 5 3 5 6 6 5 4 4 6 5 6 Footnote: *All laboratory tests are plasma concentrations or activities unless stated otherwise. **Each test is classified as quantitative (QN) or qualitative (QL). ***The number of decision levels specified is given for quantitative variables and the number of options specified for qualitative variables. Alarm/Alert Systems Variables specified in the Alarms/Alerts System. Test* Type** Categories*** Critical Care Chemistry Sub-system: Bilirubin (mol/l) Blood prothrombin time Albumin (g/l) Total protein (g/l) Aspartate aminotransferase (u/l) Amylase (u/l) Calcium (mmol/l) Phosphate (mmol/l) Creatine kinase (u/l) Alkaline phosphatase (u/l) Alanine aminotransferase (u/l) Gamma-glutamyl transferase (u/l) Urate (mmol/l) Triglyceride (mmol/l) Blood haemoglobin concentration Blood white cell count Magnesium (mmol/l) QN QN QN QN QN QN QN QN QN QN QN QN QN QN QN QN QN 15 5 14 13 15 5 20 19 15 17 15 15 4 18 8 7 6 Variables considered at the Integrating System level: Renal Diagnostic Outcome Acid-Base Diagnostic Outcome Haemodynamic Diagnostic Outcome Critical Care Chemistry Diagnostic Outcome QL QL QL QL Coded Coded Coded Coded Variables available to all participating systems: Age Sex Clinical information QN QL QL Adult, Child M, F Coded entries A-Z entries entries entries entries 000-099 000-099 000-099 000-099 Footnote: *All laboratory tests are plasma concentrations or activities unless stated otherwise. **Each test is classified as quantitative (QN) or qualitative (QL). ***The number of decision levels specified is given for quantitative variables and the number of options specified for qualitative variables. Alarm/Alert Systems Example of output of the alarms/alerts System. 60 year old male with pyrexia (39 Celsius) PlasmaSodium Potassium Urea Creatinine 152 3.9 4.5 98 Diagnostic Outcomes Renal DSS: Acid-Base DSS: Haemodynamic DSS: Univariate Alert: Moderate hypernatraemia. No Results. Univariate Alert: Pyrexia Critical Care Chemistry DSS: Integrating Interpretation Integrating KBS: mmol/l[135-145] mmol/l[3.5-5.0] mmol/l[3.0-7.0] mol/l [60-100] No Results. Multivariate Alert: Dehydration in a pyrexial patient. Decision Support Systems Support for Health Care Professionals • Microbiology Applications – Computerised infectious disease monitor (Evans, 1986). Computer-generated alarms were produced for (1) all patients with hospital-acquired infections; (2) patients on antibiotics to which they were not susceptible; (3) who could be receiving less expensive antibiotics; (4) who were receiving prophylactic antibiotcis for too long. Use of the system saved time for hospital infection control staff, and improved antibiotic use. – MRSA monitor (Safran, Scherrer 1994). Infection control nurses were provided daily with computer-generated lab alerts giving details of MRSA+ together with re-admission alerts giving details of new admissions previously known to be colonised with MRSA. System saved time for ICNs and helped as a preventive warning. Decision Support Systems Support for Health Care Professionals • Histopathology/Cytopathology applications – PAPNET Cervical Screening Neural Network – Telepathology Decision Support Systems Support for Health Care Professionals • Haematology applications – Diamond and Nguyen/Coulter Electronics. Various DSS applications for use with Coulter systems, flow cytometry, classification of haematological malignancies etc. EQA Applications • EQA Toolkit