Optimize for Time and Space

Verily's wearable sensor pipeline needs to find, for each patient, every pair of heart-rate readings taken within 5 minutes of each other where the readings differ by more than 30 bpm, since that pattern can indicate a sensor fault worth flagging for review. Here's the function a colleague wrote for a single patient's sorted readings.

import java.util.ArrayList;
import java.util.List;

public class SensorFaultDetector {

    public List<int[]> findSuspiciousPairs(List<Reading> readings) {
        List<int[]> flagged = new ArrayList<>();
        for (int i = 0; i < readings.size(); i++) {
            for (int j = 0; j < readings.size(); j++) {
                Reading a = readings.get(i);
                Reading b = readings.get(j);
                long minutesApart = Math.abs(a.timestampMillis - b.timestampMillis) / 60000;
                int bpmDiff = Math.abs(a.bpm - b.bpm);
                if (minutesApart <= 5 && bpmDiff > 30) {
                    flagged.add(new int[]{i, j});
                }
            }
        }
        return flagged;
    }

    public static class Reading {
        long timestampMillis;
        int bpm;

        public Reading(long timestampMillis, int bpm) {
            this.timestampMillis = timestampMillis;
            this.bpm = bpm;
        }
    }
}

readings is already sorted by timestamp in ascending order. Optimize this function for time and space complexity, and explain the complexity of your version compared to the original, given that patients can have tens of thousands of readings from a multi-day wearable session.