What Causes the Top 100 Chronic Illnesses in America: Part 2

How Much Chronic Illness Do Our Drugs Cause? A data-driven estimate — and a call for iatrogenic vigilance by physicians, policymakers, and patients.

Mar 20, 2026

As always, nothing in this article constitutes medical advice. Consult with your physician before starting or stopping any medication, or changing dose. This is a medical practice critique and policy document. Please share with your physician, Congressional Representatives, your Senators, and your friends and family. Together, we can cut the annual cost of medicine and, more importantly, reduce human pain and suffering through knowledge. -JLW, PhD

Chronic illness costs the United States an estimated $4.9 trillion annually (93% of total) — a figure that has risen sharply in recent years. It accounts for 90 percent of all healthcare spending. It drives nearly every major public health crisis the country faces. Part 1 of this series laid out the 100 conditions at the center of that burden; together, they are responsible for around 90% of all healthcare cost in America.

This article is an objective estimate of about how much of it we are doing to ourselves — through the very drugs prescribed to treat it.

If medications contribute to even 10 percent of that $4.9 trillion burden, that is a nearly $500 billion problem. For context: that is more than the entire annual budget of the Department of Defense. It is more than the US spends on Medicare prescription drug coverage. And unlike aging, obesity, or genetics, it is modifiable. Today. Without discovering a single new drug.

~$490 billion. Attributable to medications. Modifiable. Today.

PATIENT NARRATIVE (composite, anonymized) Margaret is 67 years old. She has had rheumatoid arthritis for a decade, managed with low-dose prednisone and long-term naproxen. In the past 18 months, her primary care physician has referred her to four specialists:

A nephrologist — for worsening kidney function (creatinine rising over two years).
A cardiologist — for new-onset heart failure with preserved ejection fraction.
A neurologist — for memory complaints and mild executive dysfunction.
A second rheumatologist — because her joint pain has been worsening despite treatment.

Each specialist has run their own workup. Combined cost so far: over $22,000. Nobody has stopped the naproxen. Nobody has connected the four presentations to the drug regimen that has been in place for six years.

Margaret is not a rare case.

We will come back to Margaret. She is the patient this analysis was built to protect.

Causes of Chronic Illness

Let us be clear about what this article is not arguing. The dominant drivers of chronic illness are not in dispute: obesity, smoking, physical inactivity, ultra-processed diet, aging, genetics, socioeconomic stress, and environmental exposures account for the overwhelming majority of the conditions in the Part 1 list. Nothing in this analysis changes that.

What this analysis identifies is a distinct, quantifiable, and — critically — modifiable contributor that is systematically underweighted in clinical practice: the chronic use of medications whose adverse effects on the very conditions they are meant to treat, or on other conditions in the same patient, are well-documented, well-evidenced, and routinely overlooked.

The argument is not that drugs are the problem. The argument is that their iatrogenic footprint is large enough to matter at the population level, identifiable and concentrated enough to be targeted, and ignored often enough that the ignoring has a name.

The Analysis: What We Did and What We Found

Margaret’s nephrologist orders a renal ultrasound, a kidney biopsy referral, and a full metabolic panel. The drug list — naproxen for six years, low-dose prednisone — is noted in the chart but not acted on. The workup costs $6,400 and finds nonspecific changes consistent with chronic nephropathy. The discharge summary reads: ‘possible hypertensive nephropathy, continue monitoring.’ A clinician with access to the analysis described below would have had NSAIDs flagged as a known cause of CKD in seconds, for free, before a single test was ordered.

Starting from the 100 top chronic illness conditions in Part 1 (expanded to 103 in this analysis), we systematically identified every pharmaceutical drug or drug class with credible published evidence of causing, worsening, or accelerating each condition. Our sources included FDA drug labeling, peer-reviewed systematic reviews, and regulatory safety communications. Every link was graded on a four-tier evidence scale. Only the two strongest tiers — established direct effects and well-supported exacerbation signals — were used in the burden estimates below. Lower-confidence associations are included in the underlying tables but clearly flagged.

The result: 103 conditions mapped to 688 drug-condition links across 324 unique drug names and drug classes.

TECHNICAL NOTE — EVIDENCE TIERING

Tier 1 — Established direct/exacerbator: Direct medication effect or strong exacerbation signal with consistent regulatory and peer-reviewed support.

Examples: NSAIDs → hypertension; glucocorticoids → hyperglycemia and type 2 diabetes; anticholinergics → cognitive impairment.

Tier 2 — Indirect/exacerbator: Worsens existing illness or drives risk factors rather than causing the condition de novo. Examples: antipsychotics → atherosclerosis (via dyslipidemia and weight gain); opioids → fibromyalgia symptom amplification (via central sensitization).

Tier 3 — Mixed/observational: Association present; causality not settled; context-dependent.

Example: PPIs → CKD (confounding by indication is a major methodological concern; PPIs are not treated as equivalent nephrotoxins to aminoglycosides or cisplatin in this analysis).

Tier 4 — No high-confidence direct signal: Condition is structural or genetic; only nonspecific exacerbators identified.

Examples: cystic fibrosis; coalworker’s pneumoconiosis.

Burden estimates below use Tiers 1 and 2 only.

To estimate what share of the overall chronic illness burden these drug-related condition links represent, we anchored the analysis to published population attributable fractions — PAFs — for specific drug-condition pairs where such data exist. A PAF answers the question: what fraction of this disease burden would be eliminated if this drug exposure were removed? We then weighted those PAFs by each condition category’s share of total chronic disease burden.

TECHNICAL NOTE — Population Attributable Fractions (PAF) ANCHORS

1. Anticholinergics → dementia/MCI: ~10% PAF (Coupland et al., JAMA Internal Medicine 2019; large UK case-control study, adjusted for confounding by indication).

2. NSAIDs → heart failure admissions: ~19% increased admission risk in current NSAID users (Arfè et al., BMJ 2016; large European analysis across 27 million patients, OR 1.19; replicated across multiple national cohorts).

3. Glucocorticoids + antipsychotics → incident type 2 diabetes: 2–10% of general-population cases; higher in directly exposed subgroups (Hwang et al.).

4. NSAIDs + nephrotoxins → CKD progression / hospital AKI: 10–15% of at-risk progression (Lapi et al.).

5. BP-raising medications → uncontrolled/resistant hypertension: ~15–18% of hypertensive patients concurrently use NSAIDs, antidepressants, or corticosteroids (Moser & Setaro; Page et al.).

6. Opioid-induced hyperalgesia + MSK drug toxicities → pain syndrome burden: incremental 5–15% exacerbation in clinically relevant subgroups.

These PAFs apply to condition categories accounting for approximately 65–75% of US chronic disease prevalence and ~90% of healthcare costs. The primary source of uncertainty is PAF transferability: whether fractions measured in high-risk study populations scale accurately to the general population. (This is reflected in the wide confidence interval below).

A methodological objection must be addressed directly before presenting the result, because experienced clinicians and alert readers will raise it immediately.

ON THE QUESTION OF ‘DOUBLE-COUNTING’

Is it legitimate to count the same patient’s drug-worsened hypertension, CKD, and heart failure as three separate burden entries? From a healthcare policy perspective, Yes. Here is why.

Total chronic disease burden — whether measured in dollars, DALYs, hospitalizations, or quality-adjusted life years — is counted in patient-condition encounters, not unique patients. A patient with three iatrogenic conditions generates three hospitalizations, three specialist referrals, three billing codes, and three separate quality-of-life impairments. The burden is real and additive.

The appropriate caution is not about counting multiple conditions in the same patient. It is about PAF transferability — whether PAFs measured in high-risk subgroups (elderly NSAID users with existing cardiovascular disease) can be scaled to the general population. That is the real uncertainty, and it is reflected in the confidence interval below.

Multiple iatrogenic conditions in the same patient is not a statistical artifact. It is the clinical reality this analysis was designed to measure. Think about it: If we had to count the same harmed patient once for the side effects or chronic illness that results from a pharmaceutical, doctors, practices and hospitals could only bill for one condition even if the patient experienced three. (NB: Perhaps to drive incentive to avoid losses this rule should be implemented!)

The result of our analysis:

9–12% of total US chronic illness burden is attributable to or significantly exacerbated by the drugs identified in this analysis. Plausible upper bound: ~15%.

Uncertainty range: ±6–8 percentage points, driven primarily by PAF transferability across exposure groups and the predominance of exacerbation rather than de novo causation in the evidence base. To be precise: the majority of the links in this analysis describe medications worsening conditions that already exist or accelerating conditions in patients who are predisposed — not conjuring disease from nothing. This distinction matters for interpreting the estimate: it measures the modifiable medication-related contribution to ongoing chronic illness burden, not a claim that 10 percent of chronic disease diagnoses would not exist without these drugs.

But care for even “just” exacerbations and “just” accelerations cost. At the 10% midpoint, that maps to approximately $490 billion in annual US healthcare expenditure — using the most current CDC figure of $4.9 trillion in total US health spending.

Two essential caveats belong here, alongside the estimate rather than at the end where they could read as retractions:

First, the net benefit of these drugs is not subtracted. Medical experts say that many patients need them. They say that glucocorticoids and calcineurin inhibitors save lives. They say opioids treat real pain. And they say that antipsychotics treat serious mental illness.

The goal of this analysis is not to condemn these medications but to quantify the cost of using them without monitoring for their known effects — and to make that cost visible.

Second, this is a first-pass estimate. It does not yet incorporate drug-drug interaction chains — where Drug A causes Condition X, which is treated with Drug B, which causes Condition Y. That is the subject of Part 3, and it is expected to raise the estimate further. So our estimate is a conservative one at best.

The Dominant Drugs — and Who Prescribes Them

Which drugs are doing the most harm? That’s a complicated question. Just like picking the top 100 conditions was based on breadth and prevalence, the ranking of drugs doing the largest % of iatrogenic pathogenesis requires careful accounting. One straightforward measure is the number of chronic illness conditions flagged.

These are not exotic oncology agents dispensed in academic medical centers. The drugs at the top of the iatrogenic footprint list are among the most widely prescribed medications in America. Some are available over the counter. The likelihood that one of them is in your medicine cabinet, or in the cabinet of someone you care about, is substantial.

Primary care and general prescribing — highest population exposure:

Glucocorticoids (26 conditions flagged). The single highest-footprint drug class in this entire analysis. Used for allergies, autoimmune disease, joint inflammation, skin conditions, COPD exacerbations, and dozens of other indications. Linked to hyperglycemia, diabetes, weight gain, hypertension, heart failure, cognitive impairment, osteoporosis, cataracts, adrenal suppression, and susceptibility to infection.
Opioids (18 conditions flagged). Beyond dependence, opioid-induced hyperalgesia — in which opioids paradoxically amplify pain sensitivity with chronic use — means these drugs can worsen the very condition they were prescribed to treat.
NSAIDs (10 conditions flagged). Available without prescription. Used by millions of Americans daily for headaches, arthritis, and back pain. Linked to hypertension, heart failure, kidney disease, GI injury, and stroke risk. This is not controversial — it is on the label.
Antipsychotics (8 conditions flagged). The second-generation agents (olanzapine, clozapine, quetiapine) carry a well-documented metabolic burden: weight gain, dyslipidemia, insulin resistance, and type 2 diabetes.
Benzodiazepines and Z-drugs (8 conditions flagged). Respiratory depression, cognitive impairment, fall risk, and dependence in long-term users. Prescribed for anxiety and sleep. Among the most frequently cited drug classes in avoidable hospitalizations in the elderly.

Next we turn to the specialist-initiated with high per-patient condition counts:

Calcineurin inhibitors (transplant and rheumatology; 12 conditions flagged). Hypertension, CKD, dyslipidemia, diabetes, neurotoxicity, and increased malignancy risk. Essential for transplant survival. Their toxicity profile requires lifelong monitoring that is not always delivered.
Lithium (psychiatry; 12 conditions flagged). CKD, nephrogenic diabetes insipidus, thyroid dysfunction, cardiac conduction changes, weight gain, cognitive effects. Effective for bipolar disorder. Its narrow therapeutic window demands monitoring that is not universally practiced.
Immune checkpoint inhibitors (oncology; 11 conditions flagged). A relatively new drug class that has transformed cancer treatment and introduced a new category of immune-mediated adverse effects: thyroid dysfunction, type 1 diabetes, inflammatory arthritis, colitis, pneumonitis, and dermatitis. The patients who benefit most are often those least equipped to recognize these effects as drug-induced.
Amiodarone (cardiology; 10 conditions flagged). Thyroid toxicity (both hypo- and hyperthyroidism), hepatotoxicity, pulmonary toxicity, peripheral neuropathy, and photosensitivity. Its half-life is measured in months. Its adverse effects can appear and persist long after the drug is discontinued.

The policy implication are clear:

(1) The drugs with the highest population exposure are largely primary care-prescribed.

(2) The drugs with the highest per-patient condition count are specialist-prescribed.

Iatrogenic vigilance is needed at both levels — but it looks different. Primary care physicians carry the breadth problem; specialists carry the depth problem. Neither group is currently held structurally accountable for the other’s prescribing. And there is a fairly simple solution: Every specialist should check their new referrals’ script list for drugs that could exacerbate, accelerate or cause the condition they are being asked to study and treat.

The National Polypharmic Epidemic and Prescribing Cascades

Margaret has now seen four specialists in 18 months. Her medication list has grown. The nephrologist added an antihypertensive for the blood pressure flagged in her workup. The cardiologist added a diuretic for her heart failure. Her primary care physician added a PPI for the gastric irritation the naproxen has been causing for years.

The naproxen is still in the regimen. The drug that initiated the cascade is still being taken twice daily. It is now buried under three medications prescribed to manage its consequences. This is called a prescribing cascade. It is not rare; it leads to polypharmacy. Nationally, people over 65 are experiencing a polypharmic epidemic.

Depending on the setting and definition used, somewhere between 40–65% of Americans over 65 experience polypharmacy, making it one of the most widespread and consequential health issues in older adults.

More than 4 out of every 10 adults aged 65 or older (over 40%) reported using 5 or more prescription medications in the past 30 days between 2017 and 2020 (Harris, 2024; JAMA Network)

About 90% of seniors 65+ take at least one medication, 42% take five or more, and at least 18% are on ten or more drugs chronically. Data shows that the proportion of older adults taking five or more medications tripled from 13.8% in 1994 to 42.4% in 2014. (Charlesworth et al., 2015)

The trend is sharply upward. Between 1988 and 2010, the median number of prescription medications used among adults 65+ doubled from 2 to 4, and the proportion taking 5 or more medications tripled from 12.8% to 39%. (Charlesworth et al., 2015; PubMed Central).

Among the elderly (65+), the polypharmacy prevalence rose from 23.5% to 44.1% between 1999–2000 and 2017–2018 — it is also considerably higher among adults with heart disease (61.7%) and diabetes (57.7%; Wang et al., 2023 PubMed Central).

In clinical settings, the numbers are even higher. Among adults 65+ seen in U.S. physician offices between 2009 and 2016 (over 2 billion patient visits), 65.1% had some form of polypharmacy — 16.2% minor (2–3 meds), 12.1% moderate (4–5 meds), and 36.8% major (more than 5 meds; (Young et al, 2021; PLOS).

Among nursing home residents specifically, up to 60% are on five or more medications. (PharmD Live)

Even though people 65+ comprise about 14% of the total U.S. population, they account for over one-third of all outpatient spending on prescription medications. (Varghese and Patel, 2024; NCBI) With roughly 58 million Americans currently over 65, the ~42% rate translates to approximately 24 million people taking 5+ medications — and that number is growing rapidly as the population ages.

It is not the result of bad doctors. It is the result of a system that does not require anyone to ask the right question first.

What follows is not an attack on physicians. The vast majority of clinicians prescribing these medications are competent, well-trained, and doing their best for their patients. The argument is structural, not personal.

The Facts Are Known, or Knowable, But Not Accessed and Used

What is known:

The iatrogenic effects catalogued in this analysis are not obscure research findings. They are in every drug label. They are taught in every pharmacology course. NSAIDs raise blood pressure and damage kidneys — it is a FDA black box warning. Glucocorticoids cause hyperglycemia — it is in every package insert. Anticholinergics impair cognition in elderly patients — it has been on the Beers Criteria for inappropriate geriatric prescribing since 1991. The knowledge exists and has existed for decades. It is not the problem.

What is practiced:

What happens in clinical practice — especially at the specialist level — is systematic disconnection. The prescriber who initiated the causative medication is often not the clinician who sees the new condition. Referral pathways are unidirectional: the cardiologist receives the patient with new heart failure and works it up as a cardiac problem. The nephrologist receives the patient with rising creatinine and works it up as a kidney problem. The drug list is reviewed, but the question — could this be drug-induced? — is not structurally required by the diagnostic workup, the referral form, or the billing code. It depends entirely on whether an individual clinician thinks to ask it. Many do not, not because they are negligent, but because the system is not designed to prompt the question.

What the standard requires:

The standard of care in medicine has long included the obligation to consider iatrogenic causation. Drug-induced disease appears in the differential diagnosis of virtually every condition in Harrison’s Principles of Internal Medicine. Medication reconciliation is a Joint Commission patient safety standard. The American Geriatrics Society explicitly lists drug review as the first step in evaluating new symptoms in older patients. This is not a new standard. It is not even in dispute. What is in dispute — or should be — is the magnitude of the gap between that standard and its application at the population level.

When the harm is foreseeable, the warning is on the label, and the question is never asked — that is not oversight. That is negligence.

Note well that prescribing or continuing these drugs without monitoring for their known iatrogenic effects, and failing to consider iatrogenic causation when those effects present clinically, is below the standard of care. When the evidence is as well-established as it is in this analysis — when the FDA has put it on the label, when the guidelines have put it in the differential, when the patient is sitting in a specialist’s office with a condition that is explicitly listed as an adverse effect of a drug she has been taking for six years — and the connection is not made, that failure has a clinical and legal name. The word is negligence. In cases of pattern, of systemic failure, of institutional disregard for a known and documented risk, the stronger word is malfeasance.

We use that word deliberately and specifically. Not to assign blame to individual physicians. To describe a structural failure in how medicine is practiced at scale. And to present a legal framework from which change — real change — can be envisioned.

The structural fix is obvious. The problem is most acute — and most correctable — at the point of specialist referral. When a primary care physician refers a patient to a cardiologist, nephrologist, or neurologist for a new chronic condition, the referral rarely includes a structured review of medications as potential causes. The specialist receives the patient as a disease to be worked up, not as a drug-exposure history to be interrogated.

The fix is not a clinical protocol. It is a structural one: referral templates that require the field ‘iatrogenic differential considered: yes/no, specify’ as a mandatory entry before the referral, and before a specialist workup begins. This requirement costs nothing to implement. It requires no new training, no new technology, and no new evidence. It requires only that the system be designed to ask the question it already knows how to answer.

And it can be automated. Patients who conditions may be being made worse by medicine should be flagged, and medicines that may be making patients sick should be flagged.

What Should Change

The following are not aspirational suggestions. They are concrete actions that can be taken today, with tools and authorities that already exist.

FOR POLICY MAKERS

1. Require iatrogenic differential documentation in specialist referral templates. Make ‘iatrogenic cause considered: yes/no’ a mandatory field for CMS-reimbursed specialist consultations.

Cost: near zero. Potential impact: redirect billions in avoidable diagnostic workup.

2. Fund structured medication review as a standalone reimbursable service for patients with 3 or more chronic conditions and 5 or more concurrent medications. This model exists in some Medicare Advantage plans. It is not universal. It should be. (Prevention of iatrogenically induced chronic illness may occur if the conditions (diagnoses) of patients who have 2 or more chronic conditions and/or 4 or more concurrent medications could be studied for potential matches to medications.

3. Correct iatrogenic attribution in chronic disease surveillance. ICD-10 code E66.1 — drug-induced obesity — is documented in the literature but almost never billed. The same is true across dozens of iatrogenic codes. Attribution drives resource allocation. Undercoding iatrogenic causes structurally depresses the measured scale of the problem. Incentivize the discovery of same by generous compensation for the prevention of harm.

4. Commission a population-level systematic review of PAFs for drug-induced chronic conditions. The estimate in this article is carefully constructed but synthesized from condition-specific studies. A purpose-built meta-analysis could produce a more precise and policy-defensible number and would almost certainly raise the central estimate.

FOR CLINICIANS — ESPECIALLY SPECIALISTS RECEIVING REFERRALS AND THOSE MAKING THEM

1. Before making a referral, medication/condition matching should be conducted. Before specialists initiate a new diagnostic workup for a chronic condition, they should require of themselves to answer one question: is this on the known adverse-effect profile of any medication this patient is taking This is not a new standard. It is the standard. It is being structurally under-applied.

2. Review the medication list as part of the differential, not as an afterthought after the workup is complete. The differential diagnosis begins before the first test is ordered.

3. Communicate backward to the referring physician. If you suspect a drug-induced etiology, say so explicitly in the consult note. Do not assume someone else will make the connection. In a fragmented referral system, if you do not write it, it does not exist.

4. Use the tools that exist. The underlying tables from this analysis are available as a structured reference. Any drug name entered into the patient risk checker returns every condition in the 103-most prevalent chronic condition set that drug is known to affect — with ICD code, evidence tier, and mechanism.

While it takes less time than ordering a basic metabolic panel, this should be made a billable service.

FOR PATIENTS

1. At every new diagnosis, or specialist appointment, before any test is ordered, ask this question:

‘Can you check to see if any of my current medications be causing or worsening what you are seeing?’

Write it down before you go in. You are entitled to ask it. If the answer is dismissive, ask for the reasoning.

2. Keep a complete and current medication list — including over-the-counter drugs, supplements, and anything prescribed by any specialist. Bring it to every appointment, every time.

The clinician who does not have your full medication list cannot make this assessment.

3. Before accepting a new prescription for a new symptom, ask: ‘Is this symptom a known side effect of anything I am already taking?’ This question can prevent the prescribing cascade — the pattern in which a drug’s side effect is treated with another drug, whose side effect is treated with another drug. Margaret’s four-specialist, $22,000 journey began with that cascade.

The Resolution

Six months after a structured medication review — initiated not by a specialist, but by Margaret herself, who had read something online and asked the question at her next primary care visit — the naproxen was discontinued. A non-pharmacologic pain management plan was initiated. The prednisone was tapered to the lowest effective dose with closer glycemic monitoring.Her creatinine stabilized over the following quarter. Her blood pressure normalized without the new antihypertensive, which was discontinued. Her cardiologist, reviewing the record at a six-month follow-up, noted that her heart failure presentation may have been drug-induced all along. The PPI was no longer needed either, and was stopped.Net medications removed: four.Net new diagnoses confirmed: zero.

Net cost of the medication review that started this: one conversation.

The tools to identify these relationships exist. The evidence base to act on them exists. The standard of care that requires asking the question exists.

To policy makers: build the iatrogenic differential into the structural machinery of healthcare — referral templates, reimbursement codes, surveillance data — and you will reduce chronic disease burden without discovering a single new drug.

To clinicians: the drug list is part of the differential. Check it first.

To patients: you are allowed to ask whether your medications are making you sick. Ask before the workup begins. Ask every time.

The question costs nothing. Not asking it costs nearly $500 billion.

Part 3 of this series will examine drug–drug interaction chains: where Drug A causes Condition X, which is treated with Drug B, which causes Condition Y. The cascade effect is expected to raise the burden estimate further. Subscription readers will be notified when it publishes.

The underlying drug-condition reference tables, evidence tier documentation, and patient risk checker tool used in this analysis are available to readers on request. Email: info@ipak-edu.org Subject Line: Dr. Lyons-Weiler’s Iatrogenic Illness/Medication Map

Discussion about this post

Ready for more?