I’m a doctor. I worry every time I prescribe painkillers to a patient.

"Please, I need my Oxycodone!" my patient, M, pleaded with me.

My eyes met his. I observed every fleeting facial expression, hoping to gauge his intentions. The discussion about whether to continue to prescribe this medication was one I'd had too many times with too many patients over the past few months.

The prescription painkiller and heroin epidemic, explained

Middle-aged white Americans are dying at faster rates, and nobody knows why

America’s harrowing painkiller and heroin epidemic, explained in five charts

"My arthritis is always worst in the winter," he said, rubbing his lower back.

It was a snowy afternoon in clinic, and M and I were in the midst of a debate. Oxycodone is an opioid medication, and, like other painkillers such as Oxycontin, Percocet, and Vicodin, it carries a significant risk of abuse.

M said he needed the pills for their pain-relieving effects. He wanted a new prescription. I was disinclined. Opioids are highly addictive. They're often abused. Worst, they decrease the body's drive to breathe, making them deadly in some cases. As much as I wanted to trust M, his story didn't quite add up. Was he abusing the drug, even selling it? Given the rising toll of prescription narcotics, these questions weren't unreasonable.

In Massachusetts where I am a physician, unintentional deaths from opioid overdoses increased from 5.3 to 10.1 per 100,000 residents between 2000 and 2013. In 2014, the number jumped to 18.6 per 100,000. These numbers include overdoses from heroin, which works the same way as opioid pills. Some people who become addicted to painkillers, unable to afford more medication or secure a prescription, then turn to heroin. But as of 2015, prescription opiates on their own account for 44 deaths each day in the United States.

In 2014, then-Massachusetts Gov. Deval Patrick declared opioid abuse and overdose a public health emergency. In June 2015, a task force established to address the issue recommended a plan that would set aside nearly $28 million to tackle the epidemic from numerous angles.

Because opioid abuse and addiction is such a widespread problem, the patients who receive prescriptions for these pills are not always the people who take them. There is a large street market for opioids, and once in the possession of people who abuse them, prescription painkillers — along with anti-anxiety medications, such as benzodiazepines like Klonopin — can become even more dangerous when incorporated into potent drug cocktails (much like cocaine-and-heroin "speedballing"). These mixtures can be lethal given the unpredictability and variability in their contents

The possibility of drug abuse, overdose, and diversion is the backdrop to every conversation I have with a patient about opioids. Some cases are clear-cut. A patient in pain from terminal cancer, whose need for narcotics is obvious and whose potential for dependence is immaterial — I don't worry too much with patients like that. But in most cases the decision €”is far more fraught.

My task as a doctor is to take stock of each patient's risk for misuse of the medicines and weight it against the desire to treat his or her pain. There is an ever-present fear that, as much as I hate to believe it, a patient could be manipulating me.

I often recall the surprise, betrayal, and alarm one of my colleagues experienced when police caught her patient selling the pain pills she'd prescribed him for years. Safeguards such as Massachusetts's prescription monitoring program, €”which logs all controlled substances prescribed to a patient and tests for drugs in the urine, €”are helpful but can still be circumvented.

But my worst fear isn't the legal possibility of supplying an addict — so long as safeguards are reasonably followed, doctors are protected from their patient's criminal behavior. What I fear most is harming a patient or, worse yet, unwittingly playing a role in someone's death.

The simplest solution to avoid these risks, of course, is to not start patients on narcotics at all, instead relying on physical therapy, non-opioid pain medicines, and other adjuncts. But patients sometimes come to me already taking opioids. I inherited M from another physician who left the practice, and when he became my patient, he was already on a relatively high dose of Oxycodone.

His previous doctor started him on the painkillers after major back surgery with the goal of weaning him off them after he had recovered. But unlike other patients with clear motives — some sought a short course of painkillers for acute pain, for example, and then stopped the medicines as soon as possible — €”M's case was tricky.

The pain test

I don't want to deny pain relief to patients who truly feel opioids help them. A prima facie refusal to ever prescribe opioids contradicts expert opinion; according to the American Pain society, for the right patients and under close monitoring, narcotics can indeed be an option as part of a chronic pain regimen.

But I do discuss the data behind narcotics for pain relief with my patients. A recent study showed that opioids in conjunction with the non-narcotic painkiller naproxen for acute lower back pain worked no better than taking naproxen alone.

In addition, the data behind narcotics for long-term management of back pain is flimsy. A patient knows his or her body best, and when it comes to certain types of pain, some may truly feel they benefit from opioid medications.

M certainly had reasons for pain, including an arthritic spine and extensive hardware in his back. But a few parts of his story didn't quite make sense. Although his urine tests had always shown he was taking the Oxycodone and no other illicit substances, he wasn't showing any symptoms of withdrawal despite running out of his high-dose Oxycodone a few weeks prior. This made me wonder whether he was actually taking all of the pills, or if he was selling some.

What I did believe was that his back pain was unrelenting, despite physical therapy and other so-called conservative measures. Ultimately, I told him it was time to transition from Oxycodone to a safer medicine, with a slow taper of the opioid to prevent symptoms of withdrawal.

Reluctantly at first, he helped me derive a plan to switch pain medicines over time. We agreed to work toward transitioning his pain regimen to non-narcotic options with the help of a specialist. After our discussion, he appeared upbeat and even expressed appreciation for my desire to treat his pain as safely as possible.

As M left my office, my mind was spinning, running through all the other options we could have chosen. I had used my judgment to do what I thought was best, but it hadn't been easy. Had he been misusing the narcotics? I will likely never know. In the end, of course, doctors can't control what patients do outside of the office.

Although the extent of this iteration of the drug epidemic is new, the decision whether to prescribe these pills ultimately rests on a principle as old as medicine itself: trust between doctor and patient, and the desire to foster and preserve it.

Allison Bond is a resident physician in internal medicine at Massachusetts General Hospital in Boston.

First Person is Vox's home for compelling, provocative narrative essays. Do you have a story to share? Read our submission guidelines, and pitch us at firstperson@vox.com.

Painkillers now kill more Americans than any illegal drug