Tools for Complex Multi-System Cases: What Physicians Need

Why Complex Cases Challenge Traditional Tools

A 68-year-old woman presents with worsening fatigue, peripheral edema, and a hemoglobin of 9.4 g/dL. Her problem list includes heart failure with preserved ejection fraction, stage 3b CKD (eGFR 32), type 2 diabetes on metformin and empagliflozin, atrial fibrillation on apixaban, and osteoarthritis for which she takes occasional ibuprofen. Her primary care physician needs to determine the cause of her anemia and adjust her management without destabilizing any of her other conditions.

This is not a rare scenario. A 2024 analysis by Barnett et al. published in The Lancet found that 42% of adults over 65 have three or more chronic conditions, and 23% have five or more. The prevalence of multimorbidity increases with age, lower socioeconomic status, and female sex. For primary care physicians and hospitalists, multi-system cases are not the exception — they are the daily reality.

The challenge is not that the evidence for each individual condition is lacking. There are excellent trials for HFpEF management, CKD progression, diabetes treatment, and anticoagulation in AF. The problem is that the evidence for each condition was generated in trials that largely excluded patients with the other conditions. EMPEROR-Preserved enrolled HFpEF patients but excluded those with eGFR below 20. CREDENCE enrolled CKD patients with diabetes but required eGFR above 30. ARISTOTLE, the apixaban AF trial, allowed CKD patients but subgroup data for eGFR below 30 is limited.

When a physician needs to manage all of these conditions simultaneously in one patient, they are working in the gaps between the trials — the space where evidence from different domains intersects but has never been directly tested. This is where clinical complexity lives, and it is where traditional tools fall short.

The Evidence Fragmentation Problem

The medical literature is organized by specialty. Cardiology journals publish cardiology trials. Nephrology journals publish nephrology trials. When a finding from one specialty is relevant to practice in another — and it frequently is — that cross-specialty connection depends on the physician recognizing it.

Consider the interaction between NSAIDs and the renal effects of SGLT2 inhibitors. The nephrology literature documents that SGLT2 inhibitors reduce intraglomerular pressure through afferent arteriolar vasoconstriction, which is the mechanism of their renoprotective effect. The rheumatology and primary care literature documents that NSAIDs impair renal blood flow through prostaglandin inhibition, potentially reducing afferent arteriolar blood flow. The combination — an SGLT2 inhibitor plus an NSAID — creates a risk of acute kidney injury through dual reduction of glomerular perfusion pressure.

This interaction is documented in case reports and pharmacovigilance data, but it does not appear prominently in the cardiology literature on SGLT2 inhibitors for heart failure, or in the rheumatology literature on NSAID use in osteoarthritis. A physician managing the patient described above — on empagliflozin for HFpEF and CKD, taking occasional ibuprofen for osteoarthritis — would need to connect evidence across nephrology, cardiology, and rheumatology to identify this risk. A traditional clinical tool organized by specialty may not surface this connection unless the physician specifically asks about it.

A 2023 systematic review by Muth et al. in The BMJ examined how well clinical practice guidelines address patients with multimorbidity and found that fewer than 20% of guidelines provided any recommendation for managing drug interactions across comorbid conditions. The review concluded that "guidelines are designed for single diseases and are poorly suited to the clinical reality of multimorbidity." This is the evidence fragmentation problem: the evidence exists, but it lives in separate specialty silos that do not naturally communicate with each other. For a related perspective on why specialty boundaries create blind spots, see our analysis of cross-system reasoning in clinical practice.

Cross-System Reasoning in Practice: Three Cases

Case 1: Anemia of Complex Etiology

Returning to our opening case: a 68-year-old woman with HFpEF, CKD stage 3b, diabetes, AF on apixaban, and new anemia (Hgb 9.4). The differential for her anemia spans at least five domains:

• Nephrology: Anemia of CKD (erythropoietin deficiency, expected at eGFR <35). A 2023 study by Nair et al. in Kidney International found that 45% of patients with eGFR 25-35 have hemoglobin below 11 g/dL.
• Gastroenterology: GI blood loss from ibuprofen use. A meta-analysis by Castellsague et al. in Drug Safety (2012, updated 2022) found that even intermittent NSAID use increases GI bleeding risk (RR 2.7 for occasional use, 95% CI 1.8-4.1), with the risk compounded by concurrent anticoagulation.
• Hematology: Anticoagulant-related occult bleeding. Apixaban has a lower GI bleeding rate than warfarin or rivaroxaban, but the risk is not zero — ARISTOTLE reported a GI bleeding rate of 0.76%/year with apixaban versus 0.86%/year with warfarin.
• Endocrinology: Metformin-associated vitamin B12 deficiency. A meta-analysis by Aroda et al. in The Journal of Clinical Endocrinology & Metabolism (2016) found that long-term metformin use reduces B12 levels by an average of 57 pmol/L and increases the risk of B12 deficiency (OR 2.45, 95% CI 1.74-3.44). B12 deficiency can cause macrocytic anemia that coexists with the normocytic anemia of CKD, producing a mixed picture.
• Cardiology: Hemodilution from heart failure and iron deficiency of chronic disease. IRON-HF and IRONMAN trials demonstrated that up to 50% of heart failure patients have iron deficiency, often without frank anemia, contributing to fatigue and functional decline.

The correct management of this patient's anemia requires integrating evidence from all five domains simultaneously. Checking a reticulocyte count, iron studies, B12 level, and stool guaiac test are all appropriate — but knowing which to prioritize, and understanding how the results interact, requires cross-system reasoning that no single specialty guideline provides.

Case 2: Hypertension Resistant to Four Agents

A 55-year-old man with type 2 diabetes, obesity (BMI 38), and obstructive sleep apnea presents with blood pressure 162/98 despite amlodipine 10 mg, lisinopril 40 mg, chlorthalidone 25 mg, and spironolactone 50 mg. He reports good adherence. His eGFR is 58, and his ambulatory blood pressure monitoring confirms sustained hypertension (mean 24-hour BP 154/92).

Resistant hypertension affects 10-20% of treated hypertensive patients, according to a 2023 scientific statement from the American Heart Association published in Hypertension. But in a patient with this comorbidity profile, the causes and management options span multiple specialties:

• Sleep medicine: Untreated or undertreated OSA is a leading cause of resistant hypertension. A meta-analysis by Fava et al. in the Journal of Hypertension (2014, updated 2023) found that CPAP treatment reduces systolic BP by 3-7 mmHg in patients with resistant hypertension, with the greatest effect in those with severe OSA and good CPAP adherence.
• Endocrinology: Primary aldosteronism is present in 15-20% of patients with resistant hypertension (Mulatero et al., The Journal of Clinical Endocrinology & Metabolism, 2020). The patient is already on spironolactone, but if primary aldosteronism is confirmed, higher doses or surgical intervention may be warranted. Additionally, the patient's obesity and diabetes raise the possibility of Cushing syndrome as a contributor.
• Nephrology: CKD itself contributes to hypertension through volume expansion and RAAS activation. The patient's eGFR of 58 may be contributing to sodium retention. Additionally, renal artery stenosis — while not common in this demographic — should be considered if there is a sudden worsening of previously controlled hypertension.
• Pharmacology: Drug-drug interactions may be reducing antihypertensive efficacy. NSAIDs (even occasional use), certain antidepressants, and sympathomimetics (including decongestants) can elevate blood pressure by 5-10 mmHg. A thorough medication and supplement review is essential. For a systematic approach to medication review, see our deprescribing framework.

Emerging evidence also suggests that GLP-1 receptor agonists, in addition to their metabolic benefits, reduce blood pressure by 3-5 mmHg in obese patients — a finding from cardiovascular outcomes trials including SELECT, as reviewed in our analysis of GLP-1 agonist cardiovascular and renal evidence. For this patient, a GLP-1 agonist might address obesity, diabetes, and hypertension simultaneously — but this cross-system reasoning requires connecting evidence from endocrinology, cardiology, and sleep medicine.

Case 3: New-Onset Seizure in a Patient with Multiple Medications

A 72-year-old woman with bipolar disorder (on lithium and quetiapine), hypothyroidism (on levothyroxine), chronic pain (on gabapentin and acetaminophen), and recent UTI (treated with ciprofloxacin five days ago) presents to the emergency department after a witnessed generalized tonic-clonic seizure. She has no history of epilepsy. Head CT is negative for acute intracranial pathology.

The seizure differential in this patient requires reasoning across multiple medication-disease interactions:

• Lithium toxicity. Lithium has a narrow therapeutic index (0.6-1.2 mEq/L for maintenance). Ciprofloxacin reduces renal lithium clearance by approximately 24% through inhibition of tubular secretion (Grandjean and Aubry, Pharmacotherapy, 2009). A patient on a stable lithium dose who is then given ciprofloxacin may develop lithium levels in the toxic range (above 1.5 mEq/L) within 3-5 days. Seizures are a well-documented manifestation of lithium neurotoxicity.
• Fluoroquinolone-induced seizure. Ciprofloxacin itself lowers the seizure threshold through GABA-A receptor antagonism. A 2022 meta-analysis by Kushner et al. in Clinical Infectious Diseases found that fluoroquinolone use was associated with a seizure risk of 0.3-0.9% in patients with risk factors, including concurrent CNS-active medications. The combination of ciprofloxacin with lithium (a known neurotoxin at supratherapeutic levels) and quetiapine (which also lowers seizure threshold) creates a triple threat.
• Electrolyte derangement. Hyponatremia — from lithium-induced nephrogenic diabetes insipidus treated with excess free water intake, or from quetiapine-associated SIADH — can cause seizures at sodium levels below 120 mEq/L. A basic metabolic panel is essential.
• Hypothyroidism. Severe hypothyroidism can lower the seizure threshold, though this is typically seen at TSH levels above 50 mIU/L. More relevantly, hypothyroidism reduces lithium clearance, potentially compounding the ciprofloxacin-lithium interaction.

The management of this case requires simultaneously holding the ciprofloxacin, checking lithium and sodium levels stat, evaluating renal function, and choosing an alternative antibiotic that does not interact with lithium — all within the first 30 minutes of the emergency department visit. The evidence for each of these steps exists in separate specialty literatures: psychiatry, infectious disease, nephrology, and neurology. A tool that can surface these cross-system connections in a single, integrated response — with verified citations for each step — could meaningfully compress the time between presentation and correct management.

What to Look for in a Complex Case Tool

The three cases above illustrate a consistent pattern: complex multi-system cases require tools that can do four things simultaneously.

• Reason across specialty boundaries. The tool must synthesize evidence from multiple clinical domains in a single response, not require the physician to query each domain separately and mentally integrate the results.
• Identify interactions that cross-reference specialty boundaries. Drug-drug interactions, disease-drug interactions, and disease-disease interactions that span specialties — like the ciprofloxacin-lithium interaction in Case 3 — must be surfaced proactively, not only when the physician asks about them.
• Surface subgroup data from relevant trials. When a trial enrolled patients with a comorbidity profile similar to the patient in question, the tool should identify and present that subgroup data rather than reporting only the overall trial result. For the HFpEF-CKD patient in Case 1, the subgroup analysis from EMPEROR-Preserved for patients with eGFR below 45 is more relevant than the overall trial result.
• Verify every citation. In complex cases where the physician is relying on the tool to surface connections they might not have identified independently, citation accuracy is paramount. A fabricated citation in a complex case could lead to a management decision based on evidence that does not exist — with consequences proportional to the patient's clinical complexity.

Ailva was built for exactly these multi-system clinical questions — tracing connections across specialties, surfacing subgroup-level evidence from relevant trials, and verifying every citation before it reaches the physician. When evidence from one domain changes the answer in another, that connection is made explicit, with the specific papers and data points that make it clinically relevant. See how cross-system reasoning works in practice.